There are a number of cases where a checksum delta is more appropriate than the default time-based delta:
* Timeline has switched since the prior backup
* File timestamp is older than recorded in the prior backup
* File size changed but timestamp did not
* File timestamp is in the future compared to the start of the backup
* Online option has changed since the prior backup
A practical example is that checksum delta will be enabled after a failover to standby due to the timeline switch. In this case, timestamps can't be trusted and our recommendation has been to run a full backup, which can impact the retention schedule and requires manual intervention.
Now, a checksum delta will be performed if the backup type is incr/diff. This means more CPU will be used during the backup but the backup size will be smaller and the retention schedule will not be impacted.
Contributed by Cynthia Shang.
Use checksums rather than timestamps to determine if files have changed. This is useful in cases where the timestamps may not be trustworthy, e.g. when performing an incremental after failing over to a standby.
If checksum delta is enabled then checksums will be used for verification of resumed backups, even if they are full. Resumes have always used checksums to verify the files in the repository, enabling delta performs checksums on the database files as well.
Note that the user must manually enable this feature in cases were it would be useful or just keep in enabled all the time. A future commit will address automatically enabling the feature in cases where it seems likely to be useful.
Contributed by Cynthia Shang.
Previously an error would be generated if other files were present and not owned by the PostgreSQL user. This hasn't been a big deal in practice but it could cause issues.
Also add tests to make sure the same logic applies with links to files, i.e. all other files in the directory should be ignored. This was actually working correctly, but there were no tests for it before.
Relative link paths were being combined with the paths of previous links (relative or absolute) due to the $strPath variable being modified in the current iteration rather than simply being passed to the next level of recursion.
This issue did not affect absolute links and relative tablespace links were caught by other checks, though the error was confusing.
Reported by Cynthia Shang.
Implemented using the same logic as the patches adding this feature to PostgreSQL, 8694cc96 and 920a5e50. Temporary relation exclusion is enabled in PostgreSQL ≥ 9.0. Unlogged relation exclusion is enabled in PostgreSQL ≥ 9.1, where the feature was introduced.
Contributed by Cynthia Shang.
* Exclude contents of pg_snapshots, pg_serial, pg_notify, and pg_dynshmem from backup since they are rebuilt on startup.
* Exclude pg_internal.init files from backup since they are rebuilt on startup.
The archive_status directory is now recreated on restore to support PostgreSQL 8.3 which does not recreate it automatically like more recent versions do.
Also fixed log checking after PostgreSQL shuts down to include FATAL messages and disallow immediate shutdowns which can throw FATAL errors in the log.
Reported by Stephen Frost.
The integration tests that were supposed to prevent this regression did not work as intended. They verified the contents of a table in the (supposedly) restored tablespace, deleted the table, and then deleted the tablespace. All of this was deemed sufficient to prove that the tablespace had been restored correctly and was valid.
However, PostgreSQL will happily recreate a tablespace on the basis of a single full-page write, at least in the affected versions. Since writes to the test table were replayed from WAL with each recovery, all the tests passed even though the tablespace was missing after the restore.
The tests have been updated to include direct comparisons against the file system and a new table that is not replayed after a restore because it is created before the backup and never modified again.
Versions ≥ 9.0 were not affected due to numerous synthetic integration tests that verify backups and restores file by file.
Refactor storage layer to allow for new repository filesystems using drivers. (Reviewed by Cynthia Shang.)
Refactor IO layer to allow for new compression formats, checksum types, and other capabilities using filters. (Reviewed by Cynthia Shang.)
* Refactor Ini.pm to facilitate testing.
* Complete statement/branch coverage for Ini.pm.
* Improved functions used to test/munge manifest and info files.
This option allows pgBackRest to validate page checksums in data files when checksums are enabled on PostgreSQL >= 9.3. Note that this functionality requires a C library which may not initially be available in OS packages. The option will automatically be enabled when the library is present and checksums are enabled on the cluster.
A connection to the primary cluster is still required to start/stop the backup and copy files that are not replicated, but the vast majority of files are copied from the standby in order to reduce load on the master.
These include (depending on the version where they were introduced): pgsql_tmp, pg_dynshmem, pg_notify, pg_replslot, pg_serial, pg_snapshots, pg_stat_tmp, pg_subtrans. The postgresql.auto.conf.tmp file is now excluded in addition to files that were already excluded: backup_label.old, postmaster.opts, postmaster.pid, recovery.conf, recovery.done.
* Tablespace paths that had $PGDATA as a substring would be identified as a subdirectories of $PGDATA even when they were not.
* Also hardened relative path checking a bit.
Improved handling of users/groups captured during backup that do not exist on the restore host. Also explicitly handle the case where user/group is not mapped to a name.
The pg_xlogfile_name() function is no longer used to construct WAL filenames from LSNs. While this function is convenient it is not available on a standby. Instead, the archive is searched for the LSN in order to find the timeline. If due to some misadventure the LSN appears on multiple timelines then an error will be thrown, whereas before this condition would have passed unnoticed.
Some files need to be added to the manifest after the initial build. This is currently done in only one place but usage will expand in the future so the functionality has been encapsulated in addFile().
Added database version constants and changed version identification code to use hash tables instead of if-else. Propagated the db version constants to the rest of the code and in passing fixed some path/filename constants.
Added new regression tests to check that specific files are never copied.
This feature can result in major space and time savings when only specific databases are restored. Unrestored databases will not be accessible but must be manually dropped before they will be removed from the shared catalogue.
* All files and directories linked from PGDATA are now included in the backup. By default links will be restored directly into PGDATA as files or directories. The --link-all option can be used to restore all links to their original locations. The --link-map option can be used to remap a link to a new location.
* Removed --tablespace option and replaced with --tablespace-map-all option which should more clearly indicate its function.
* Added detail log level which will output more information than info without being as verbose as debug.
* 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.
