1
0
mirror of https://github.com/pgbackrest/pgbackrest.git synced 2024-12-14 10:13:05 +02:00
pgbackrest/lib/pgBackRest/Backup.pm

926 lines
39 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.
2014-03-06 03:53:13 +03:00
####################################################################################################################################
# BACKUP MODULE
####################################################################################################################################
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
package pgBackRest::Backup;
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.
2014-03-06 03:53:13 +03:00
use strict;
use warnings FATAL => qw(all);
use Carp qw(confess);
use English '-no_match_vars';
2014-07-28 01:13:23 +03:00
use Exporter qw(import);
use Fcntl 'SEEK_CUR';
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.
2014-03-06 03:53:13 +03:00
use File::Basename;
use File::Path qw(remove_tree);
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
use pgBackRest::Common::Exception;
use pgBackRest::Common::Exit;
use pgBackRest::Common::Ini;
use pgBackRest::Common::Log;
use pgBackRest::Common::Wait;
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
use pgBackRest::Archive;
use pgBackRest::ArchiveCommon;
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
use pgBackRest::BackupCommon;
use pgBackRest::BackupFile;
use pgBackRest::BackupInfo;
use pgBackRest::Common::String;
use pgBackRest::Config::Config;
use pgBackRest::Db;
use pgBackRest::DbVersion;
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
use pgBackRest::File;
use pgBackRest::FileCommon;
use pgBackRest::Manifest;
use pgBackRest::Protocol::Common;
use pgBackRest::Protocol::LocalProcess;
use pgBackRest::Protocol::Protocol;
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
use pgBackRest::Version;
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.
2014-03-06 03:53:13 +03:00
####################################################################################################################################
# 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.
2014-03-06 03:53:13 +03:00
####################################################################################################################################
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.
2014-03-06 03:53:13 +03:00
{
my $class = shift; # Class name
# Create the class hash
my $self = {};
bless $self, $class;
# Assign function parameters, defaults, and log debug info
my ($strOperation) = logDebugParam(__PACKAGE__ . '->new');
# Return from function and log return values if any
return logDebugReturn
(
$strOperation,
{name => 'self', value => $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.
2014-03-06 03:53:13 +03:00
}
####################################################################################################################################
# fileNotInManifest
#
# Find all files in a backup path that are not in the supplied manifest.
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.
2014-03-06 03:53:13 +03:00
####################################################################################################################################
sub fileNotInManifest
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.
2014-03-06 03:53:13 +03:00
{
my $self = shift;
# Assign function parameters, defaults, and log debug info
my
(
$strOperation,
$oFileLocal,
$strPathType,
$oManifest,
$oAbortedManifest
) =
logDebugParam
(
__PACKAGE__ . '->fileNotInManifest', \@_,
{name => 'oFileLocal', trace => true},
{name => 'strPathType', trace => true},
{name => 'oManifest', trace => true},
{name => 'oAbortedManifest', trace => true}
);
# Build manifest for aborted temp path
my $hFile = $oFileLocal->manifest($strPathType);
# Get compress flag
my $bCompressed = $oAbortedManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS);
2014-09-14 22:31:56 +03:00
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.
2014-03-06 03:53:13 +03:00
my @stryFile;
foreach my $strName (sort(keys(%{$hFile})))
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.
2014-03-06 03:53:13 +03:00
{
# Ignore certain files that will never be in the manifest
if ($strName eq FILE_MANIFEST ||
$strName eq '.')
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.
2014-03-06 03:53:13 +03:00
{
next;
}
# Get the file type (all links will be deleted since they are easy to recreate)
my $cType = $hFile->{$strName}{type};
# If a directory check if it exists in the new manifest
if ($cType eq 'd')
{
if ($oManifest->test(MANIFEST_SECTION_TARGET_PATH, $strName))
{
next;
}
}
# Else if a file
elsif ($cType eq 'f')
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.
2014-03-06 03:53:13 +03:00
{
# If the original backup was compressed the remove the extension before checking the manifest
my $strFile = $strName;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
if ($bCompressed)
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.
2014-03-06 03:53:13 +03:00
{
$strFile = substr($strFile, 0, length($strFile) - 3);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
}
# To be preserved the file must exist in the new manifest and not be a reference to a previous backup
if ($oManifest->test(MANIFEST_SECTION_TARGET_FILE, $strFile) &&
!$oManifest->test(MANIFEST_SECTION_TARGET_FILE, $strFile, MANIFEST_SUBKEY_REFERENCE))
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.
2014-03-06 03:53:13 +03:00
{
# To be preserved the checksum must be defined
my $strChecksum = $oAbortedManifest->get(MANIFEST_SECTION_TARGET_FILE, $strFile, MANIFEST_SUBKEY_CHECKSUM, false);
# The timestamp should also match and the size if the file is not compressed. If the file is compressed it's
# not worth extracting the size - it will be hashed later to verify its authenticity.
if (defined($strChecksum) &&
($bCompressed || ($oManifest->numericGet(MANIFEST_SECTION_TARGET_FILE, $strFile, MANIFEST_SUBKEY_SIZE) ==
$hFile->{$strName}{size})) &&
$oManifest->numericGet(MANIFEST_SECTION_TARGET_FILE, $strFile, MANIFEST_SUBKEY_TIMESTAMP) ==
$hFile->{$strName}{modification_time})
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.
2014-03-06 03:53:13 +03:00
{
$oManifest->set(MANIFEST_SECTION_TARGET_FILE, $strFile, MANIFEST_SUBKEY_CHECKSUM, $strChecksum);
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.
2014-03-06 03:53:13 +03:00
next;
}
}
}
# Push the file/path/link to be deleted into the result array
push @stryFile, $strName;
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.
2014-03-06 03:53:13 +03:00
}
# Return from function and log return values if any
return logDebugReturn
(
$strOperation,
{name => 'stryFile', value => \@stryFile}
);
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.
2014-03-06 03:53:13 +03:00
}
####################################################################################################################################
# tmpClean
2014-06-04 18:58:30 +03:00
#
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.
2014-03-06 03:53:13 +03:00
# Cleans the temp directory from a previous failed backup so it can be reused
####################################################################################################################################
sub tmpClean
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.
2014-03-06 03:53:13 +03:00
{
my $self = shift;
# Assign function parameters, defaults, and log debug info
my
(
$strOperation,
$oFileLocal,
$oManifest,
$oAbortedManifest
) =
logDebugParam
(
__PACKAGE__ . '->tmpClean', \@_,
{name => 'oFileLocal', trace => true},
{name => 'oManifest', trace => true},
{name => 'oAbortedManifest', trace => true}
);
&log(DETAIL, 'clean backup temp path: ' . $oFileLocal->pathGet(PATH_BACKUP_TMP));
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.
2014-03-06 03:53:13 +03:00
# Get the list of files that should be deleted from temp
my @stryFile = $self->fileNotInManifest($oFileLocal, PATH_BACKUP_TMP, $oManifest, $oAbortedManifest);
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.
2014-03-06 03:53:13 +03:00
foreach my $strFile (sort {$b cmp $a} @stryFile)
{
my $strDelete = $oFileLocal->pathGet(PATH_BACKUP_TMP, $strFile);
2014-06-04 18:58:30 +03:00
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.
2014-03-06 03:53:13 +03:00
# If a path then delete it, all the files should have already been deleted since we are going in reverse order
if (!-X $strDelete && -d $strDelete)
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
{
logDebugMisc($strOperation, "remove path ${strDelete}");
rmdir($strDelete)
or confess &log(ERROR, "unable to delete path ${strDelete}, is it empty?", ERROR_PATH_REMOVE);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
}
# Else delete a file
else
{
logDebugMisc($strOperation, "remove file ${strDelete}");
fileRemove($strDelete);
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.
2014-03-06 03:53:13 +03:00
}
}
# Return from function and log return values if any
return logDebugReturn($strOperation);
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.
2014-03-06 03:53:13 +03:00
}
####################################################################################################################################
# processManifest
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.
2014-03-06 03:53:13 +03:00
#
# Process the file level backup. Uses the information in the manifest to determine which files need to be copied. Directories
# and tablespace links are only created when needed, except in the case of a full backup or if hardlinks are requested.
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.
2014-03-06 03:53:13 +03:00
####################################################################################################################################
sub processManifest
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.
2014-03-06 03:53:13 +03:00
{
my $self = shift;
# Assign function parameters, defaults, and log debug info
my
(
$strOperation,
$oFileMaster,
$strDbMasterPath,
$strDbCopyPath,
$strType,
$strDbVersion,
$bCompress,
$bHardLink,
$oBackupManifest # Manifest for the current backup
) =
logDebugParam
(
__PACKAGE__ . '->processManifest', \@_,
{name => 'oFileMaster'},
{name => 'strDbMasterPath'},
{name => 'strDbCopyPath'},
{name => 'strType'},
{name => 'strDbVersion'},
{name => 'bCompress'},
{name => 'bHardLink'},
{name => 'oBackupManifest'},
);
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.
2014-03-06 03:53:13 +03:00
2016-09-06 15:35:02 +02:00
# Start backup test point
&log(TEST, TEST_BACKUP_START);
# Get the master protocol for keep-alive
my $oProtocolMaster = protocolGet(DB, $self->{iMasterRemoteIdx});
$oProtocolMaster->noOp();
# Initialize the backup process
my $oBackupProcess = new pgBackRest::Protocol::LocalProcess(DB);
2016-09-06 15:35:02 +02:00
if ($self->{iCopyRemoteIdx} != $self->{iMasterRemoteIdx})
{
$oBackupProcess->hostAdd($self->{iMasterRemoteIdx}, 1);
}
$oBackupProcess->hostAdd($self->{iCopyRemoteIdx}, optionGet(OPTION_PROCESS_MAX));
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.
2014-03-06 03:53:13 +03:00
# Variables used for parallel copy
my $lFileTotal = 0;
my $lSizeTotal = 0;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
# If this is a full backup or hard-linked then create all paths and tablespace links
if ($bHardLink || $strType eq BACKUP_TYPE_FULL)
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.
2014-03-06 03:53:13 +03:00
{
# Create paths
foreach my $strPath ($oBackupManifest->keys(MANIFEST_SECTION_TARGET_PATH))
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.
2014-03-06 03:53:13 +03:00
{
2016-09-06 15:35:02 +02:00
$oFileMaster->pathCreate(PATH_BACKUP_TMP, $strPath);
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.
2014-03-06 03:53:13 +03:00
}
if (optionGet(OPTION_REPO_LINK))
{
for my $strTarget ($oBackupManifest->keys(MANIFEST_SECTION_BACKUP_TARGET))
{
if ($oBackupManifest->isTargetTablespace($strTarget))
{
$oFileMaster->linkCreate(
PATH_BACKUP_TMP, $strTarget, PATH_BACKUP_TMP, MANIFEST_TARGET_PGDATA . "/${strTarget}", false, true);
}
}
}
}
# Iterate all files in the manifest
foreach my $strRepoFile (
2016-09-06 15:35:02 +02:00
sort {sprintf("%016d-${b}", $oBackupManifest->numericGet(MANIFEST_SECTION_TARGET_FILE, $b, MANIFEST_SUBKEY_SIZE)) cmp
sprintf("%016d-${a}", $oBackupManifest->numericGet(MANIFEST_SECTION_TARGET_FILE, $a, MANIFEST_SUBKEY_SIZE))}
($oBackupManifest->keys(MANIFEST_SECTION_TARGET_FILE, 'none')))
{
# If the file has a reference it does not need to be copied since it can be retrieved from the referenced backup.
# However, if hard-linking is turned on the link will need to be created
my $strReference = $oBackupManifest->get(MANIFEST_SECTION_TARGET_FILE, $strRepoFile, MANIFEST_SUBKEY_REFERENCE, false);
if (defined($strReference))
{
# If hardlinking is turned on then create a hardlink for files that have not changed since the last backup
if ($bHardLink)
{
logDebugMisc($strOperation, "hardlink ${strRepoFile} to ${strReference}");
$oFileMaster->linkCreate(
PATH_BACKUP_CLUSTER, "${strReference}/${strRepoFile}", PATH_BACKUP_TMP, "${strRepoFile}", true, false, true);
}
2016-09-06 15:35:02 +02:00
# Else log the reference
else
{
logDebugMisc($strOperation, "reference ${strRepoFile} to ${strReference}");
}
2016-09-06 15:35:02 +02:00
# This file will not need to be copied
next;
}
# By default put everything into a single queue
my $strQueueKey = MANIFEST_TARGET_PGDATA;
# If the file belongs in a tablespace then put in a tablespace-specific queue
if (index($strRepoFile, DB_PATH_PGTBLSPC . '/') == 0)
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.
2014-03-06 03:53:13 +03:00
{
$strQueueKey = DB_PATH_PGTBLSPC . '/' . (split('\/', $strRepoFile))[1];
}
# Create the file hash
my $bIgnoreMissing = true;
my $strDbFile = $oBackupManifest->dbPathGet($strDbCopyPath, $strRepoFile);
my $iHostConfigIdx = $self->{iCopyRemoteIdx};
# Certain files must be copied from the master
if ($oBackupManifest->boolGet(MANIFEST_SECTION_TARGET_FILE, $strRepoFile, MANIFEST_SUBKEY_MASTER))
{
$strDbFile = $oBackupManifest->dbPathGet($strDbMasterPath, $strRepoFile);
$iHostConfigIdx = $self->{iMasterRemoteIdx};
}
# Make sure that pg_control is not removed during the backup
if ($strRepoFile eq MANIFEST_TARGET_PGDATA . '/' . DB_FILE_PGCONTROL)
{
$bIgnoreMissing = false;
}
# Increment file total and size
my $lSize = $oBackupManifest->numericGet(MANIFEST_SECTION_TARGET_FILE, $strRepoFile, MANIFEST_SUBKEY_SIZE);
$lFileTotal++;
$lSizeTotal += $lSize;
# Queue for parallel backup
$oBackupProcess->queueJob(
$iHostConfigIdx, $strQueueKey, $strRepoFile, OP_BACKUP_FILE,
[$strDbFile, $strRepoFile, $lSize,
$oBackupManifest->get(MANIFEST_SECTION_TARGET_FILE, $strRepoFile, MANIFEST_SUBKEY_CHECKSUM, false), $bCompress,
$oBackupManifest->numericGet(MANIFEST_SECTION_TARGET_FILE, $strRepoFile, MANIFEST_SUBKEY_TIMESTAMP, false),
$bIgnoreMissing]);
# Size and checksum will be removed and then verified later as a sanity check
$oBackupManifest->remove(MANIFEST_SECTION_TARGET_FILE, $strRepoFile, MANIFEST_SUBKEY_SIZE);
$oBackupManifest->remove(MANIFEST_SECTION_TARGET_FILE, $strRepoFile, MANIFEST_SUBKEY_CHECKSUM);
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
}
# pg_control should always be in the backup (unless this is an offline backup)
if (!$oBackupManifest->test(MANIFEST_SECTION_TARGET_FILE, MANIFEST_FILE_PGCONTROL) && optionGet(OPTION_ONLINE))
{
confess &log(ERROR, DB_FILE_PGCONTROL . " must be present in all online backups\n" .
'HINT: is something wrong with the clock or filesystem timestamps?', ERROR_FILE_MISSING);
}
2016-09-06 15:35:02 +02:00
# If there are no files to backup then we'll exit with an error unless in test mode. The other way this could happen is if
# the database is down and backup is called with --no-online twice in a row.
if ($lFileTotal == 0 && !optionGet(OPTION_TEST))
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.
2014-03-06 03:53:13 +03:00
{
confess &log(ERROR, "no files have changed since the last backup - this seems unlikely", ERROR_FILE_MISSING);
}
# Running total of bytes copied
my $lSizeCurrent = 0;
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.
2014-03-06 03:53:13 +03:00
# Determine how often the manifest will be saved
my $lManifestSaveCurrent = 0;
my $lManifestSaveSize = int($lSizeTotal / 100);
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.
2014-03-06 03:53:13 +03:00
if (optionSource(OPTION_MANIFEST_SAVE_THRESHOLD) ne SOURCE_DEFAULT ||
$lManifestSaveSize < optionGet(OPTION_MANIFEST_SAVE_THRESHOLD))
{
$lManifestSaveSize = optionGet(OPTION_MANIFEST_SAVE_THRESHOLD);
}
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.
2014-03-06 03:53:13 +03:00
# Run the backup jobs and process results
while (my $hyJob = $oBackupProcess->process())
{
foreach my $hJob (@{$hyJob})
{
($lSizeCurrent, $lManifestSaveCurrent) = backupManifestUpdate(
$oBackupManifest, optionGet(optionIndex(OPTION_DB_HOST, $hJob->{iHostConfigIdx}), false), $hJob->{iProcessId},
@{$hJob->{rParam}}[0..3], @{$hJob->{rResult}},
$lSizeTotal, $lSizeCurrent, $lManifestSaveSize, $lManifestSaveCurrent);
}
# A keep-alive is required here because if there are a large number of resumed files that need to be checksummed
# then the remote might timeout while waiting for a command.
$oProtocolMaster->keepAlive();
}
2016-09-06 15:35:02 +02:00
# Validate the manifest
$oBackupManifest->validate();
# Return from function and log return values if any
return logDebugReturn
(
$strOperation,
{name => 'lSizeTotal', value => $lSizeTotal}
);
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.
2014-03-06 03:53:13 +03:00
}
####################################################################################################################################
# process
2014-06-04 18:58:30 +03:00
#
# Process the database backup.
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.
2014-03-06 03:53:13 +03:00
####################################################################################################################################
sub process
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.
2014-03-06 03:53:13 +03:00
{
my $self = shift;
# Assign function parameters, defaults, and log debug info
my ($strOperation) = logDebugParam(__PACKAGE__ . '->process');
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.
2014-03-06 03:53:13 +03:00
# Record timestamp start
my $lTimestampStart = time();
# Initialize the local file object
my $oFileLocal = new pgBackRest::File
(
optionGet(OPTION_STANZA),
optionGet(OPTION_REPO_PATH),
protocolGet(NONE)
);
# Store local type, compress, and hardlink options since they can be modified by the process
my $strType = optionGet(OPTION_TYPE);
my $bCompress = optionGet(OPTION_COMPRESS);
my $bHardLink = optionGet(OPTION_HARDLINK);
# Create the cluster backup and history path
$oFileLocal->pathCreate(PATH_BACKUP_CLUSTER, PATH_BACKUP_HISTORY, undef, true, 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.
2014-03-06 03:53:13 +03:00
# Load or build backup.info
my $oBackupInfo = new pgBackRest::BackupInfo($oFileLocal->pathGet(PATH_BACKUP_CLUSTER));
2015-01-20 21:42:22 +02:00
# Build backup tmp and config
my $strBackupTmpPath = $oFileLocal->pathGet(PATH_BACKUP_TMP);
my $strBackupConfFile = $oFileLocal->pathGet(PATH_BACKUP_TMP, 'backup.manifest');
2015-01-20 21:42:22 +02:00
# Declare the backup manifest
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
my $oBackupManifest = new pgBackRest::Manifest($strBackupConfFile, false);
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.
2014-03-06 03:53:13 +03:00
# Find the previous backup based on the type
my $oLastManifest;
my $strBackupLastPath;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
if ($strType ne BACKUP_TYPE_FULL)
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.
2014-03-06 03:53:13 +03:00
{
$strBackupLastPath = $oBackupInfo->last($strType eq BACKUP_TYPE_DIFF ? BACKUP_TYPE_FULL : BACKUP_TYPE_INCR);
if (defined($strBackupLastPath))
{
$oLastManifest = new pgBackRest::Manifest(
$oFileLocal->pathGet(PATH_BACKUP_CLUSTER, "${strBackupLastPath}/" . FILE_MANIFEST));
&log(INFO, 'last backup label = ' . $oLastManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_LABEL) .
', version = ' . $oLastManifest->get(INI_SECTION_BACKREST, INI_KEY_VERSION));
# If this is incr or diff warn if certain options have changed
my $strKey;
# Warn if compress option changed
if (!$oLastManifest->boolTest(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS, undef, $bCompress))
{
&log(WARN, "${strType} backup cannot alter compress option to '" . boolFormat($bCompress) .
"', reset to value in ${strBackupLastPath}");
$bCompress = $oLastManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS);
}
# Warn if hardlink option changed
if (!$oLastManifest->boolTest(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_HARDLINK, undef, $bHardLink))
{
&log(WARN, "${strType} backup cannot alter hardlink option to '" . boolFormat($bHardLink) .
"', reset to value in ${strBackupLastPath}");
$bHardLink = $oLastManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_HARDLINK);
}
}
else
{
&log(WARN, "no prior backup exists, ${strType} backup has been changed to full");
$strType = BACKUP_TYPE_FULL;
}
}
2015-01-21 01:00:03 +02:00
# Backup settings
$oBackupManifest->set(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_TYPE, undef, $strType);
$oBackupManifest->numericSet(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_TIMESTAMP_START, undef, $lTimestampStart);
$oBackupManifest->boolSet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_BACKUP_STANDBY, undef, optionGet(OPTION_BACKUP_STANDBY));
$oBackupManifest->boolSet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS, undef, $bCompress);
$oBackupManifest->boolSet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_HARDLINK, undef, $bHardLink);
$oBackupManifest->boolSet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_ONLINE, undef, optionGet(OPTION_ONLINE));
$oBackupManifest->boolSet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_ARCHIVE_COPY, undef,
!optionGet(OPTION_ONLINE) ||
(optionGet(OPTION_BACKUP_ARCHIVE_CHECK) && optionGet(OPTION_BACKUP_ARCHIVE_COPY)));
$oBackupManifest->boolSet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_ARCHIVE_CHECK, undef,
!optionGet(OPTION_ONLINE) || optionGet(OPTION_BACKUP_ARCHIVE_CHECK));
# Initialize database objects
my $oDbMaster = undef;
my $oDbStandby = undef;
# Get the database objects
($oDbMaster, $self->{iMasterRemoteIdx}, $oDbStandby, $self->{iCopyRemoteIdx}) = dbObjectGet();
# If remote copy was not explicitly set then set it equal to master
if (!defined($self->{iCopyRemoteIdx}))
{
$self->{iCopyRemoteIdx} = $self->{iMasterRemoteIdx};
}
# Initialize the master file object
my $oFileMaster = new pgBackRest::File
(
optionGet(OPTION_STANZA),
optionGet(OPTION_REPO_PATH),
protocolGet(DB, $self->{iMasterRemoteIdx})
);
# Determine the database paths
my $strDbMasterPath = optionGet(optionIndex(OPTION_DB_PATH, $self->{iMasterRemoteIdx}));
my $strDbCopyPath = optionGet(optionIndex(OPTION_DB_PATH, $self->{iCopyRemoteIdx}));
# Database info
my ($strDbVersion, $iControlVersion, $iCatalogVersion, $ullDbSysId) = $oDbMaster->info();
my $iDbHistoryId = $oBackupInfo->check($strDbVersion, $iControlVersion, $iCatalogVersion, $ullDbSysId);
$oBackupManifest->numericSet(MANIFEST_SECTION_BACKUP_DB, MANIFEST_KEY_DB_ID, undef, $iDbHistoryId);
$oBackupManifest->set(MANIFEST_SECTION_BACKUP_DB, MANIFEST_KEY_DB_VERSION, undef, $strDbVersion);
$oBackupManifest->numericSet(MANIFEST_SECTION_BACKUP_DB, MANIFEST_KEY_CONTROL, undef, $iControlVersion);
$oBackupManifest->numericSet(MANIFEST_SECTION_BACKUP_DB, MANIFEST_KEY_CATALOG, undef, $iCatalogVersion);
$oBackupManifest->numericSet(MANIFEST_SECTION_BACKUP_DB, MANIFEST_KEY_SYSTEM_ID, undef, $ullDbSysId);
# Backup from standby can only be used on PostgreSQL >= 9.1
if (optionGet(OPTION_ONLINE) && optionGet(OPTION_BACKUP_STANDBY) && $strDbVersion < PG_VERSION_BACKUP_STANDBY)
{
confess &log(
ERROR,
'option \'' . OPTION_BACKUP_STANDBY . '\' not valid for PostgreSQL < ' . PG_VERSION_BACKUP_STANDBY, ERROR_CONFIG);
}
# Start backup (unless --no-online is set)
my $strArchiveStart = undef;
my $strLsnStart = undef;
my $hTablespaceMap = undef;
my $hDatabaseMap = undef;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
# Don't start the backup but do check if PostgreSQL is running
if (!optionGet(OPTION_ONLINE))
{
if ($oFileMaster->exists(PATH_DB_ABSOLUTE, $strDbMasterPath . '/' . DB_FILE_POSTMASTERPID))
{
if (optionGet(OPTION_FORCE))
{
&log(WARN, '--no-online passed and ' . DB_FILE_POSTMASTERPID . ' exists but --force was passed so backup will ' .
2014-12-19 00:05:06 +02:00
'continue though it looks like the postmaster is running and the backup will probably not be ' .
'consistent');
}
else
{
confess &log(ERROR, '--no-online passed but ' . DB_FILE_POSTMASTERPID . ' exists - looks like the postmaster is ' .
'running. Shutdown the postmaster and try again, or use --force.', ERROR_POSTMASTER_RUNNING);
}
}
}
# Else start the backup normally
else
{
# Start the backup
($strArchiveStart, $strLsnStart) =
$oDbMaster->backupStart(
BACKREST_NAME . ' backup started at ' . timestampFormat(undef, $lTimestampStart), optionGet(OPTION_START_FAST));
2015-01-21 01:00:03 +02:00
# Record the archive start location
$oBackupManifest->set(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_ARCHIVE_START, undef, $strArchiveStart);
$oBackupManifest->set(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_LSN_START, undef, $strLsnStart);
&log(INFO, "backup start archive = ${strArchiveStart}, lsn = ${strLsnStart}");
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.
2014-03-06 03:53:13 +03:00
# Get tablespace map
$hTablespaceMap = $oDbMaster->tablespaceMapGet();
# Get database map
$hDatabaseMap = $oDbMaster->databaseMapGet();
# Wait for replay on the standby to catch up
if (optionGet(OPTION_BACKUP_STANDBY))
{
my ($strStandbyDbVersion, $iStandbyControlVersion, $iStandbyCatalogVersion, $ullStandbyDbSysId) = $oDbStandby->info();
$oBackupInfo->check($strStandbyDbVersion, $iStandbyControlVersion, $iStandbyCatalogVersion, $ullStandbyDbSysId);
$oDbStandby->configValidate();
&log(INFO, "wait for replay on the standby to reach ${strLsnStart}");
my ($strReplayedLSN, $strCheckpointLSN) = $oDbStandby->replayWait($strLsnStart);
&log(
INFO,
"replay on the standby reached ${strReplayedLSN}" .
(defined($strCheckpointLSN) ? ", checkpoint ${strCheckpointLSN}" : ''));
2016-09-06 15:35:02 +02:00
# The standby db object won't be used anymore so undef it to catch any subsequent references
undef($oDbStandby);
protocolDestroy(DB, $self->{iCopyRemoteIdx}, true);
}
}
# Build the manifest
$oBackupManifest->build($oFileMaster, $strDbVersion, $strDbMasterPath, $oLastManifest, optionGet(OPTION_ONLINE),
$hTablespaceMap, $hDatabaseMap);
2014-07-28 01:13:23 +03:00
&log(TEST, TEST_MANIFEST_BUILD);
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.
2014-03-06 03:53:13 +03:00
# Check if an aborted backup exists for this stanza
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.
2014-03-06 03:53:13 +03:00
if (-e $strBackupTmpPath)
{
my $bUsable;
my $strReason = "resume is disabled";
my $oAbortedManifest;
# Attempt to read the manifest file in the aborted backup to see if it can be used. If any error at all occurs then the
# backup will be considered unusable and a resume will not be attempted.
if (optionGet(OPTION_RESUME))
{
$strReason = "unable to read ${strBackupTmpPath}/backup.manifest";
eval
{
# Load the aborted manifest
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
$oAbortedManifest = new pgBackRest::Manifest("${strBackupTmpPath}/backup.manifest");
# Key and values that do not match
my $strKey;
my $strValueNew;
my $strValueAborted;
# Check version
if ($oBackupManifest->get(INI_SECTION_BACKREST, INI_KEY_VERSION) ne
$oAbortedManifest->get(INI_SECTION_BACKREST, INI_KEY_VERSION))
2014-10-14 22:44:50 +03:00
{
$strKey = INI_KEY_VERSION;
$strValueNew = $oBackupManifest->get(INI_SECTION_BACKREST, INI_KEY_VERSION);
$strValueAborted = $oAbortedManifest->get(INI_SECTION_BACKREST, INI_KEY_VERSION);
}
# Check format
elsif ($oBackupManifest->get(INI_SECTION_BACKREST, INI_KEY_FORMAT) ne
$oAbortedManifest->get(INI_SECTION_BACKREST, INI_KEY_FORMAT))
{
$strKey = INI_KEY_FORMAT;
$strValueNew = $oBackupManifest->get(INI_SECTION_BACKREST, INI_KEY_FORMAT);
$strValueAborted = $oAbortedManifest->get(INI_SECTION_BACKREST, INI_KEY_FORMAT);
}
# Check backup type
elsif ($oBackupManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_TYPE) ne
$oAbortedManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_TYPE))
{
$strKey = MANIFEST_KEY_TYPE;
$strValueNew = $oBackupManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_TYPE);
$strValueAborted = $oAbortedManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_TYPE);
}
# Check prior label
elsif ($oBackupManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_PRIOR, undef, false, '<undef>') ne
$oAbortedManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_PRIOR, undef, false, '<undef>'))
{
$strKey = MANIFEST_KEY_PRIOR;
$strValueNew = $oBackupManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_PRIOR, undef, false, '<undef>');
$strValueAborted = $oAbortedManifest->get(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_PRIOR, undef, false, '<undef>');
}
# Check compression
elsif ($oBackupManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS) ne
$oAbortedManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS))
{
$strKey = MANIFEST_KEY_COMPRESS;
$strValueNew = $oBackupManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS);
$strValueAborted = $oAbortedManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_COMPRESS);
2014-10-14 22:44:50 +03:00
}
# Check hardlink
elsif ($oBackupManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_HARDLINK) ne
$oAbortedManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_HARDLINK))
{
$strKey = MANIFEST_KEY_HARDLINK;
$strValueNew = $oBackupManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_HARDLINK);
$strValueAborted = $oAbortedManifest->boolGet(MANIFEST_SECTION_BACKUP_OPTION, MANIFEST_KEY_HARDLINK);
}
# If key is defined then something didn't match
if (defined($strKey))
{
$strReason = "new ${strKey} '${strValueNew}' does not match aborted ${strKey} '${strValueAborted}'";
}
# Else the backup can be resumed
else
{
$bUsable = true;
}
return true;
}
or do
{
$bUsable = false;
}
}
2014-06-04 18:58:30 +03:00
# If the aborted backup is usable then clean it
if ($bUsable)
{
&log(WARN, 'aborted backup of same type exists, will be cleaned to remove invalid files and resumed');
&log(TEST, TEST_BACKUP_RESUME);
# Clean the old backup tmp path
$self->tmpClean($oFileLocal, $oBackupManifest, $oAbortedManifest);
}
# Else remove it
else
{
&log(WARN, "aborted backup exists, but cannot be resumed (${strReason}) - will be dropped and recreated");
&log(TEST, TEST_BACKUP_NORESUME);
remove_tree($oFileLocal->pathGet(PATH_BACKUP_TMP))
or confess &log(ERROR, "unable to delete tmp path: ${strBackupTmpPath}");
$oFileLocal->pathCreate(PATH_BACKUP_TMP, undef, undef, false, true);
}
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
}
# Else create the backup tmp path
else
{
logDebugMisc($strOperation, "create temp backup path ${strBackupTmpPath}");
$oFileLocal->pathCreate(PATH_BACKUP_TMP, undef, undef, false, 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.
2014-03-06 03:53:13 +03:00
}
2015-01-20 21:42:22 +02:00
# Save the backup manifest
$oBackupManifest->save();
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.
2014-03-06 03:53:13 +03:00
# Perform the backup
my $lBackupSizeTotal =
$self->processManifest(
$oFileMaster, $strDbMasterPath, $strDbCopyPath, $strType, $strDbVersion, $bCompress, $bHardLink, $oBackupManifest);
&log(INFO, "${strType} backup size = " . fileSizeFormat($lBackupSizeTotal));
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
# Master file object no longer needed
undef($oFileMaster);
# Stop backup (unless --no-online is set)
my $strArchiveStop = undef;
my $strLsnStop = undef;
v0.10: Backup and archiving are functional This version has been put into production at Resonate, so it does work, but there are a number of major caveats. * No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery. * Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby. * Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% threadsafe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly. * Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything. * The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text. * Absolutely no documentation (outside the code). Well, excepting these release notes. * Lots of other little things and not so little things. Much refactoring to follow.
2014-03-06 03:53:13 +03:00
if (optionGet(OPTION_ONLINE))
{
($strArchiveStop, $strLsnStop, my $strTimestampDbStop, my $oFileHash) = $oDbMaster->backupStop();
2015-01-21 01:00:03 +02:00
$oBackupManifest->set(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_ARCHIVE_STOP, undef, $strArchiveStop);
$oBackupManifest->set(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_LSN_STOP, undef, $strLsnStop);
&log(INFO, "backup stop archive = ${strArchiveStop}, lsn = ${strLsnStop}");
# Write out files returned from stop backup
foreach my $strFile (sort(keys(%{$oFileHash})))
{
# Only save the file if it has content
if (defined($$oFileHash{$strFile}))
{
my $strFileName = $oFileLocal->pathGet(PATH_BACKUP_TMP, $strFile);
# Write content out to a file
fileStringWrite($strFileName, $$oFileHash{$strFile});
# Compress if required
if ($bCompress)
{
$oFileLocal->compress(PATH_BACKUP_ABSOLUTE, $strFileName);
$strFileName .= '.' . $oFileLocal->{strCompressExtension};
}
# Add file to manifest
$oBackupManifest->fileAdd(
$strFile,
(fileStat($strFileName))->mtime,
length($$oFileHash{$strFile}),
$oFileLocal->hash(PATH_BACKUP_ABSOLUTE, $strFileName, $bCompress), true);
&log(DETAIL, "wrote '${strFile}' file returned from pg_stop_backup()");
}
}
}
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.
2014-03-06 03:53:13 +03:00
# Remotes no longer needed (destroy them here so they don't timeout)
&log(TEST, TEST_BACKUP_STOP);
undef($oDbMaster);
protocolDestroy(undef, undef, true);
# If archive logs are required to complete the backup, then check them. This is the default, but can be overridden if the
# archive logs are going to a different server. Be careful of this option because there is no way to verify that the backup
# will be consistent - at least not here.
if (optionGet(OPTION_ONLINE) && optionGet(OPTION_BACKUP_ARCHIVE_CHECK))
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.
2014-03-06 03:53:13 +03:00
{
2015-01-20 21:42:22 +02:00
# Save the backup manifest a second time - before getting archive logs in case that fails
$oBackupManifest->save();
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.
2014-03-06 03:53:13 +03:00
# Create the modification time for the archive logs
my $lModificationTime = time();
# After the backup has been stopped, need to make a copy of the archive logs to make the db consistent
logDebugMisc($strOperation, "retrieve archive logs ${strArchiveStart}:${strArchiveStop}");
New simpler configuration and consistent project/exe/path naming. * The repo-path option now always refers to the repository where backups and archive are stored, whether local or remote, so the repo-remote-path option has been removed. The new spool-path option can be used to define a location for queueing WAL segments when archiving asynchronously. Otherwise, a local repository is no longer required. * Implemented a new config format which should be far simpler to use. See the User Guide and Configuration Reference for details but for a simple configuration all options can now be placed in the stanza section. Options that are shared between stanzas can be placed in the [global] section. More complex configurations can still make use of command sections though this should be a rare use case. * The default configuration filename is now pgbackrest.conf instead of pg_backrest.conf. This was done for consistency with other naming changes but also to prevent old config files from being loaded accidentally. * The default repository name was changed from /var/lib/backup to /var/lib/pgbackrest. * Lock files are now stored in /tmp/pgbackrest by default. These days /run/pgbackrest would be the preferred location but that would require init scripts which are not part of this release. The lock-path option can be used to configure the lock directory. * Log files are now stored in /var/log/pgbackrest by default and no longer have the date appended so they can be managed with logrotate. The log-path option can be used to configure the lock directory. * Executable filename changed from pg_backrest to pgbackrest.
2016-04-14 15:30:54 +02:00
my $oArchive = new pgBackRest::Archive();
my $strArchiveId = $oArchive->getArchiveId($oFileLocal);
my @stryArchive = lsnFileRange($strLsnStart, $strLsnStop, $strDbVersion);
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.
2014-03-06 03:53:13 +03:00
foreach my $strArchive (@stryArchive)
{
my $strArchiveFile =
$oArchive->walFileName($oFileLocal, $strArchiveId, $strArchive, false, optionGet(OPTION_ARCHIVE_TIMEOUT));
$strArchive = substr($strArchiveFile, 0, 24);
if (optionGet(OPTION_BACKUP_ARCHIVE_COPY))
{
logDebugMisc($strOperation, "archive: ${strArchive} (${strArchiveFile})");
# Copy the log file from the archive repo to the backup
my $strDestinationFile = MANIFEST_TARGET_PGDATA . "/pg_xlog/${strArchive}" .
($bCompress ? ".$oFileLocal->{strCompressExtension}" : '');
my $bArchiveCompressed = $strArchiveFile =~ "^.*\.$oFileLocal->{strCompressExtension}\$";
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.
2014-03-06 03:53:13 +03:00
my ($bCopyResult, $strCopyChecksum, $lCopySize) =
$oFileLocal->copy(PATH_BACKUP_ARCHIVE, "${strArchiveId}/${strArchiveFile}",
PATH_BACKUP_TMP, $strDestinationFile,
$bArchiveCompressed, $bCompress,
undef, $lModificationTime, undef, true);
# Add the archive file to the manifest so it can be part of the restore and checked in validation
my $strPathLog = MANIFEST_TARGET_PGDATA . '/pg_xlog';
my $strFileLog = "${strPathLog}/${strArchive}";
# Compare the checksum against the one already in the archive log name
if ($strArchiveFile !~ "^${strArchive}-${strCopyChecksum}(\\.$oFileLocal->{strCompressExtension}){0,1}\$")
{
confess &log(ERROR, "error copying WAL segment '${strArchiveFile}' to backup - checksum recorded with " .
"file does not match actual checksum of '${strCopyChecksum}'", ERROR_CHECKSUM);
}
# Add file to manifest
$oBackupManifest->fileAdd($strFileLog, $lModificationTime, $lCopySize, $strCopyChecksum, 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.
2014-03-06 03:53:13 +03:00
}
}
# Create the path for the new backup
my $lTimestampStop = time();
my $strBackupLabel = backupLabelFormat($strType, $strBackupLastPath, $lTimestampStop);
# Make sure that the timestamp has not already been used by a prior backup. This is unlikely for online backups since there is
# already a wait after the manifest is built but it's still possible if the remote and local systems don't have synchronized
# clocks. In practice this is most useful for making offline testing faster since it allows the wait after manifest build to
# be skipped by dealing with any backup label collisions here.
if (fileList($oFileLocal->pathGet(PATH_BACKUP_CLUSTER),
($strType eq BACKUP_TYPE_FULL ? '^' : '_') .
timestampFileFormat(undef, $lTimestampStop) .
($strType eq BACKUP_TYPE_FULL ? 'F' : '(D|I)$')) ||
fileList($oFileLocal->pathGet(PATH_BACKUP_CLUSTER, PATH_BACKUP_HISTORY . '/' . timestampFormat('%4d', $lTimestampStop)),
($strType eq BACKUP_TYPE_FULL ? '^' : '_') .
timestampFileFormat(undef, $lTimestampStop) .
($strType eq BACKUP_TYPE_FULL ? 'F' : '(D|I)\.manifest\.' . $oFileLocal->{strCompressExtension}), undef, true))
{
waitRemainder();
$strBackupLabel = backupLabelFormat($strType, $strBackupLastPath, time());
}
# Record timestamp stop in the config
$oBackupManifest->set(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_TIMESTAMP_STOP, undef, $lTimestampStop + 0);
$oBackupManifest->set(MANIFEST_SECTION_BACKUP, MANIFEST_KEY_LABEL, undef, $strBackupLabel);
# Final save of the backup manifest
2015-01-20 21:42:22 +02:00
$oBackupManifest->save();
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.
2014-03-06 03:53:13 +03:00
&log(INFO, "new backup label = ${strBackupLabel}");
# Make a compressed copy of the manifest for history
$oFileLocal->copy(PATH_BACKUP_TMP, FILE_MANIFEST,
PATH_BACKUP_TMP, FILE_MANIFEST . '.gz',
undef, true);
# Move the backup tmp path to complete the backup
logDebugMisc($strOperation, "move ${strBackupTmpPath} to " . $oFileLocal->pathGet(PATH_BACKUP_CLUSTER, $strBackupLabel));
$oFileLocal->move(PATH_BACKUP_TMP, undef, PATH_BACKUP_CLUSTER, $strBackupLabel);
# Copy manifest to history
$oFileLocal->move(PATH_BACKUP_CLUSTER, "${strBackupLabel}/" . FILE_MANIFEST . '.gz',
PATH_BACKUP_CLUSTER, PATH_BACKUP_HISTORY . qw{/} . substr($strBackupLabel, 0, 4) .
"/${strBackupLabel}.manifest.gz", true);
# Create a link to the most recent backup
$oFileLocal->remove(PATH_BACKUP_CLUSTER, LINK_LATEST);
if (optionGet(OPTION_REPO_LINK))
{
$oFileLocal->linkCreate(PATH_BACKUP_CLUSTER, $strBackupLabel, PATH_BACKUP_CLUSTER, LINK_LATEST, undef, true);
}
# Save backup info
$oBackupInfo->add($oBackupManifest);
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.
2014-03-06 03:53:13 +03:00
# Return from function and log return values if any
return logDebugReturn($strOperation);
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.
2014-03-06 03:53:13 +03:00
}
2014-06-04 18:58:30 +03:00
1;