1
0
mirror of https://github.com/pgbackrest/pgbackrest.git synced 2025-12-21 23:47:33 +02:00
Files
pgbackrest/lib/BackRest/Db.pm

516 lines
19 KiB
Perl
Raw Normal View History

v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
# DB MODULE
####################################################################################################################################
package BackRest::Db;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
use strict;
use warnings FATAL => qw(all);
use Carp qw(confess);
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
use DBD::Pg ':async';
use DBI;
use Exporter qw(import);
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
our @EXPORT = qw();
use Fcntl qw(O_RDONLY);
use File::Basename qw(dirname);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
use lib dirname($0);
use BackRest::Config;
use BackRest::Exception;
use BackRest::File;
use BackRest::Utility;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
# Operation constants
####################################################################################################################################
use constant OP_DB => 'Db';
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
use constant OP_DB_NEW => OP_DB . "->new";
use constant OP_DB_BACKUP_START => OP_DB . "->backupStart";
use constant OP_DB_BACKUP_STOP => OP_DB . "->backupStop";
use constant OP_DB_EXECUTE_SQL => OP_DB . "->executeSql";
push @EXPORT, qw(OP_DB_EXECUTE_SQL);
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
use constant OP_DB_INFO => OP_DB . "->info";
push @EXPORT, qw(OP_DB_INFO);
use constant OP_DB_TABLESPACE_MAP_GET => OP_DB . "->tablespaceMapGet";
use constant OP_DB_VERSION_GET => OP_DB . "->versionGet";
####################################################################################################################################
# Postmaster process Id file
####################################################################################################################################
use constant FILE_POSTMASTER_PID => 'postmaster.pid';
push @EXPORT, qw(FILE_POSTMASTER_PID);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
####################################################################################################################################
# Backup advisory lock
####################################################################################################################################
use constant DB_BACKUP_ADVISORY_LOCK => '12340078987004321';
push @EXPORT, qw(DB_BACKUP_ADVISORY_LOCK);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
# CONSTRUCTOR
####################################################################################################################################
sub new
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
my $class = shift; # Class name
# Create the class hash
my $self = {};
bless $self, $class;
return $self;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
####################################################################################################################################
# DESTRUCTOR
####################################################################################################################################
sub DESTROY
{
my $self = shift;
if (defined($self->{hDb}))
{
$self->{hDb}->disconnect();
undef($self->{hDb});
}
}
####################################################################################################################################
# versionSupport
#
# Returns an array of the supported Postgres versions.
####################################################################################################################################
sub versionSupport
{
my @strySupportVersion = ('8.3', '8.4', '9.0', '9.1', '9.2', '9.3', '9.4', '9.5');
return \@strySupportVersion;
}
push @EXPORT, qw(versionSupport);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
# executeSql
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
sub executeSql
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
my $self = shift;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
my $strSql = shift;
my $bIgnoreError = shift;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
logDebug(OP_DB_EXECUTE_SQL, DEBUG_CALL, undef, {isRemote => optionRemoteTypeTest(DB), script => \$strSql});
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
# Get the user-defined command for psql
my $strResult;
# Run remotely
if (optionRemoteTypeTest(DB))
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
# Build param hash
my %oParamHash;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$oParamHash{'script'} = $strSql;
if (defined($bIgnoreError) && $bIgnoreError)
{
$oParamHash{'ignore-error'} = true;
}
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
# Execute the command
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$strResult = protocolGet()->cmdExecute(OP_DB_EXECUTE_SQL, \%oParamHash, true);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
# Else run locally
else
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
if (!defined($self->{hDb}))
{
# Connect to the db
my $strDbName = 'postgres';
my $strDbUser = getpwuid($<);
my $strDbSocketPath = optionGet(OPTION_DB_SOCKET_PATH, false);
if (defined($strDbSocketPath) && $strDbSocketPath !~ /^\//)
{
confess &log(ERROR, "'${strDbSocketPath}' is not valid for '" . OPTION_DB_SOCKET_PATH . "' option:" .
" path must be absolute", ERROR_OPTION_INVALID_VALUE);
}
my $strDbUri = "dbi:Pg:dbname=${strDbName};port=" . optionGet(OPTION_DB_PORT) .
(optionTest(OPTION_DB_SOCKET_PATH) ? ';host=' . optionGet(OPTION_DB_SOCKET_PATH) : '');
logDebug(OP_DB_EXECUTE_SQL, DEBUG_MISC, 'db connect', {strDbUri => $strDbUri, strDbUser => $strDbUser});
$self->{hDb} = DBI->connect($strDbUri, $strDbUser, undef, {AutoCommit => 1, RaiseError => 0, PrintError => 0});
if (!$self->{hDb})
{
confess &log(ERROR, $DBI::errstr, ERROR_DB_CONNECT);
}
}
# Prepare the query
my $hStatement = $self->{hDb}->prepare($strSql, {pg_async => PG_ASYNC})
or confess &log(ERROR, $DBI::errstr, ERROR_DB_QUERY);
# Execute the query
$hStatement->execute();
# Wait for the query to return
my $iWaitSeconds = optionGet(OPTION_DB_TIMEOUT);
my $oExecuteWait = waitInit($iWaitSeconds);
my $bTimeout = true;
do
{
# Is the statement done?
if ($hStatement->pg_ready())
{
if (!$hStatement->pg_result())
{
# Return if the error should be ignored
if (defined($bIgnoreError) && $bIgnoreError)
{
return '';
}
# Else report it
confess &log(ERROR, $DBI::errstr . ":\n${strSql}", ERROR_DB_QUERY);
}
# Get rows and return them
my @stryArray;
do
{
@stryArray = $hStatement->fetchrow_array;
if (!@stryArray && $hStatement->err)
{
confess &log(ERROR, $DBI::errstr . ":\n${strSql}", ERROR_DB_QUERY);
}
$strResult = (defined($strResult) ? "${strResult}\n" : '') . join("\t", @stryArray);
}
while (@stryArray);
$bTimeout = false;
}
} while ($bTimeout && waitMore($oExecuteWait));
# If timeout then cancel the query and confess
if ($bTimeout)
{
$hStatement->pg_cancel();
confess &log(ERROR, "statement timed out after ${iWaitSeconds} second(s):\n${strSql}", ERROR_DB_TIMEOUT);
}
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
logDebug(OP_DB_EXECUTE_SQL, DEBUG_RESULT, undef, {strResult => $strResult});
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
return $strResult;
}
####################################################################################################################################
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# executeSqlRow
####################################################################################################################################
sub executeSqlRow
{
my $self = shift;
my $strSql = shift;
return split("\t", trim($self->executeSql($strSql)));
}
####################################################################################################################################
# executeSqlOne
####################################################################################################################################
sub executeSqlOne
{
my $self = shift;
my $strSql = shift;
return ($self->executeSqlRow($strSql))[0];
}
####################################################################################################################################
# tablespaceMapGet
#
# Get the mapping between oid and tablespace name.
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
sub tablespaceMapGet
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
my $self = shift;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
logTrace(OP_DB_TABLESPACE_MAP_GET, DEBUG_CALL);
my $oHashRef = {};
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
data_hash_build($oHashRef, "oid\tname\n" . $self->executeSql(
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
'select oid, spcname from pg_tablespace'), "\t");
return $oHashRef;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
####################################################################################################################################
# info
####################################################################################################################################
sub info
{
my $self = shift;
my $oFile = shift;
my $strDbPath = shift;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
logDebug(OP_DB_INFO, DEBUG_CALL, undef, {isRemote => $oFile->is_remote(PATH_DB_ABSOLUTE), dbPath => \$strDbPath});
# Return data from the cache if it exists
if (defined($self->{info}{$strDbPath}))
{
return $self->{info}{$strDbPath}{fDbVersion},
$self->{info}{$strDbPath}{iControlVersion},
$self->{info}{$strDbPath}{iCatalogVersion},
$self->{info}{$strDbPath}{ullDbSysId};
}
# Database info
my $iCatalogVersion;
my $iControlVersion;
my $ullDbSysId;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
my $fDbVersion;
if ($oFile->is_remote(PATH_DB_ABSOLUTE))
{
# Build param hash
my %oParamHash;
$oParamHash{'db-path'} = ${strDbPath};
# Output remote trace info
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
&log(TRACE, OP_DB_INFO . ": remote (" . $oFile->{oProtocol}->commandParamString(\%oParamHash) . ')');
# Execute the command
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
my $strResult = $oFile->{oProtocol}->cmdExecute(OP_DB_INFO, \%oParamHash, true);
# Split the result into return values
my @stryToken = split(/\t/, $strResult);
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = $stryToken[0];
$iControlVersion = $stryToken[1];
$iCatalogVersion = $stryToken[2];
$ullDbSysId = $stryToken[3];
}
else
{
# Open the control file
my $strControlFile = "${strDbPath}/global/pg_control";
my $hFile;
my $tBlock;
sysopen($hFile, $strControlFile, O_RDONLY)
or confess &log(ERROR, "unable to open ${strControlFile}", ERROR_FILE_OPEN);
# Read system identifier
sysread($hFile, $tBlock, 8) == 8
or confess &log(ERROR, "unable to read database system identifier");
$ullDbSysId = unpack('Q', $tBlock);
# Read control version
sysread($hFile, $tBlock, 4) == 4
or confess &log(ERROR, "unable to read control version");
$iControlVersion = unpack('L', $tBlock);
# Read catalog version
sysread($hFile, $tBlock, 4) == 4
or confess &log(ERROR, "unable to read catalog version");
$iCatalogVersion = unpack('L', $tBlock);
# Close the control file
close($hFile);
# Make sure the control version is supported
if ($iControlVersion == 942 && $iCatalogVersion == 201409291)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '9.4';
}
# Leave 9.5 catalog version out until it stabilizes (then move 9.5 to the top of the list)
elsif ($iControlVersion == 942) # && $iCatalogVersion == 201505311)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '9.5';
}
elsif ($iControlVersion == 937 && $iCatalogVersion == 201306121)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '9.3';
}
elsif ($iControlVersion == 922 && $iCatalogVersion == 201204301)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '9.2';
}
elsif ($iControlVersion == 903 && $iCatalogVersion == 201105231)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '9.1';
}
elsif ($iControlVersion == 903 && $iCatalogVersion == 201008051)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '9.0';
}
elsif ($iControlVersion == 843 && $iCatalogVersion == 200904091)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '8.4';
}
elsif ($iControlVersion == 833 && $iCatalogVersion == 200711281)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '8.3';
}
elsif ($iControlVersion == 822 && $iCatalogVersion == 200611241)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '8.2';
}
elsif ($iControlVersion == 812 && $iCatalogVersion == 200510211)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '8.1';
}
elsif ($iControlVersion == 74 && $iCatalogVersion == 200411041)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
$fDbVersion = '8.0';
}
else
{
confess &log(ERROR, "unexpected control version = ${iControlVersion} and catalog version = ${iCatalogVersion}" .
' (unsupported PostgreSQL version?)',
ERROR_VERSION_NOT_SUPPORTED);
}
}
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# Store data in the cache
$self->{info}{$strDbPath}{fDbVersion} = $fDbVersion;
$self->{info}{$strDbPath}{iControlVersion} = $iControlVersion;
$self->{info}{$strDbPath}{iCatalogVersion} = $iCatalogVersion;
$self->{info}{$strDbPath}{ullDbSysId} = $ullDbSysId;
&log(DEBUG, OP_DB_INFO . "=>: dbVersion = ${fDbVersion}, controlVersion = ${iControlVersion}" .
", catalogVersion = ${iCatalogVersion}, dbSysId = ${ullDbSysId}");
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
return $fDbVersion, $iControlVersion, $iCatalogVersion, $ullDbSysId;
}
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
# versionGet
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
sub versionGet
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
my $self = shift;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# Get data from the cache if possible
if (defined($self->{fVersion}) && defined($self->{strDbPath}))
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
return $self->{fVersion}, $self->{strDbPath};
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# Get version and db-path from
($self->{fVersion}, $self->{strDbPath}) =
$self->executeSqlRow("select (regexp_matches(split_part(version(), ' ', 2), '^[0-9]+\.[0-9]+'))[1], setting" .
" from pg_settings where name = 'data_directory'");
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
my $strVersionSupport = versionSupport();
if ($self->{fVersion} < ${$strVersionSupport}[0])
{
confess &log(ERROR, "unsupported Postgres version ${$strVersionSupport}[0]", ERROR_VERSION_NOT_SUPPORTED);
}
logDebug(OP_DB_VERSION_GET, DEBUG_RESULT, {dbVersion => $self->{fVersion}});
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
return $self->{fVersion}, $self->{strDbPath};
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
####################################################################################################################################
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# backupStart
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
sub backupStart
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
my $self = shift;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
my $oFile = shift;
my $strDbPath = shift;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
my $strLabel = shift;
my $bStartFast = shift;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
logDebug(OP_DB_BACKUP_START, DEBUG_CALL, undef, {dbPath => \$strDbPath, strLabel => \$strLabel, isStartFast => $bStartFast});
# Get the version from the control file
my ($fDbVersion) = $self->info($oFile, $strDbPath);
# Get version and db path from the database
my ($fCompareDbVersion, $strCompareDbPath) = $self->versionGet();
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# Error if they are not identical
if (!($fDbVersion == $fCompareDbVersion && $strDbPath eq $strCompareDbPath))
{
confess &log(ERROR, "version '${fCompareDbVersion}' and db-path '${strCompareDbPath}' queried from cluster does not match" .
" version '${fDbVersion}' and db-path '${strDbPath}' read from '${strDbPath}/global/pg_control'\n" .
"HINT: the db-path and db-port settings likely reference different clusters", ERROR_DB_MISMATCH);
}
# Only allow start-fast option for version >= 8.4
if ($self->{fVersion} < 8.4 && $bStartFast)
{
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
&log(WARN, OPTION_START_FAST . ' option is only available in PostgreSQL >= 8.4');
$bStartFast = false;
}
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# Acquire the backup advisory lock
if (!$self->executeSqlOne('select pg_try_advisory_lock(' . DB_BACKUP_ADVISORY_LOCK . ')'))
{
confess &log(ERROR, "unable to acquire backup lock\n" .
'HINT: is another backup already running on this cluster?', ERROR_LOCK_ACQUIRE);
}
# Test if a backup is already running when version >= 9.3
if (optionGet(OPTION_STOP_AUTO))
{
if ($self->{fVersion} >= 9.3)
{
if ($self->executeSqlOne('select pg_is_in_backup()'))
{
&log(WARN, 'the cluster is already in backup mode but no backup process is running. pg_stop_backup() will be called' .
' so a new backup can be started.');
$self->backupStop();
}
}
else
{
&log(WARN, OPTION_STOP_AUTO . 'option is only available in PostgreSQL >= 9.3');
}
}
&log(INFO, "executing pg_start_backup() with label \"${strLabel}\": backup will begin after " .
($bStartFast ? "the requested immediate checkpoint" : "the next regular checkpoint") . " completes");
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
my ($strTimestampDbStart, $strArchiveStart) =
$self->executeSqlRow("select to_char(current_timestamp, 'YYYY-MM-DD HH24:MI:SS.US TZ'), " .
"pg_xlogfile_name(xlog) from pg_start_backup('${strLabel}'" .
($bStartFast ? ', true' : '') . ') as xlog');
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
return $strArchiveStart, $strTimestampDbStart;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
####################################################################################################################################
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
# backupStop
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
####################################################################################################################################
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
sub backupStop
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
{
my $self = shift;
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
logDebug(OP_DB_BACKUP_STOP, DEBUG_CALL);
&log(INFO, 'executing pg_stop_backup() and waiting for all WAL segments to be archived');
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
my ($strTimestampDbStop, $strArchiveStop) =
$self->executeSqlRow("select to_char(clock_timestamp(), 'YYYY-MM-DD HH24:MI:SS.US TZ')," .
" pg_xlogfile_name(xlog) from pg_stop_backup() as xlog");
v0.80: DBI Support, Stability, and Convenience Features * Fixed an issue that caused the formatted timestamp for both the oldest and newest backups to be reported as the current time by the info command. Only text output was affected -- json output reported the correct epoch values. Reported by Michael Renner. * Fixed protocol issue that was preventing ssh errors (especially on connection) from being logged. * Now using Perl DBI and DBD::Pg for connections to PostgreSQL rather than psql. The cmd-psql and cmd-psql-option settings have been removed and replaced with db-port and db-socket-path. * Add stop-auto option to allow failed backups to automatically be stopped when a new backup starts. * Add db-timeout option to limit the amount of time pgBackRest will wait for pg_start_backup() and pg_stop_backup() to return. * Remove pg_control file at the beginning of the restore and copy it back at the very end. This prevents the possibility that a partial restore can be started by PostgreSQL. * The repository is now created and updated with consistent directory and file modes. By default umask is set to 0000 but this can be disabled with the neutral-umask setting. * Added checks to be sure the db-path setting is consistent with db-port by comparing the data_directory as reported by the cluster against the db-path setting and the version as reported by the cluster against the value read from pg_control. The db-socket-path setting is checked to be sure it is an absolute path. * Experimental support for PostgreSQL 9.5 alpha1. This may break when the control version or WAL magic changes in future versions but will be updated in each pgBackRest release to keep pace. All regression tests pass except for --target-resume tests (this functionality has changed in 9.5) and there is no testing yet for .partial WAL segments. * Major refactoring of the protocol layer to support future development. * Added vagrant test configurations for Ubuntu 14.04 and CentOS 7. * Split most of README.md out into USERGUIDE.md and CHANGELOG.md because it was becoming unwieldy. Changed most references to "database" in the user guide to "database cluster" for clarity.
2015-08-09 11:20:43 -04:00
return $strArchiveStop, $strTimestampDbStop;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-05 19:53:13 -05:00
}
1;