I am currently using pg_dump piped to gzip piped to split. But the problem with this is that all output files are always changed. So checksum-based backup always copies all data.
Are there any other good ways to perform an incremental backup of a PostgreSQL database, where a full database can be restored from the backup data?
For instance, if pg_dump could make everything absolutely ordered, so all changes are applied only at the end of the dump, or similar.
Update: Check out Barman for an easier way to set up WAL archiving for backup.
You can use PostgreSQL's continuous WAL archiving method. First you need to set wal_level=archive, then do a full filesystem-level backup (between issuing pg_start_backup() and pg_stop_backup() commands) and then just copy over newer WAL files by configuring the archive_command option.
Advantages:
Incremental, the WAL archives include everything necessary to restore the current state of the database
Almost no overhead, copying WAL files is cheap
You can restore the database at any point in time (this feature is called PITR, or point-in-time recovery)
Disadvantages:
More complicated to set up than pg_dump
The full backup will be much larger than a pg_dump because all internal table structures and indexes are included
Does not work well for write-heavy databases, since recovery will take a long time.
There are some tools such as pitrtools and omnipitr that can simplify setting up and restoring these configurations. But I haven't used them myself.
Also check out http://www.pgbackrest.org
pgBackrest is another backup tool for PostgreSQL which you should be evaluating as it supports:
parallel backup (tested to scale almost linearly up to 32 cores but can probably go much farther..)
compressed-at-rest backups
incremental and differential (compressed!) backups
streaming compression (data is compressed only once at the source and then transferred across the network and stored)
parallel, delta restore (ability to update an older copy to the latest)
Fully supports tablespaces
Backup rotation and archive expiration
Ability to resume backups which failed for some reason
etc, etc..
Another method is to backup to plain text and use rdiff to create incremental diffs.
Related
I understand that Cloud SQL ( Postgres) on-demand backup is incremental. when you restore an instance using this backup, the existing data is wiped off before the instance is restored with all new data. in other words, the "backup" process is incremental but there is no way to restore only a specific incremental backup into an instance
Please, can you confirm if the aforementioned understanding is right?
Indeed, Cloud SQL backups are incremental. Taken from the Documentation:
Backups for Second Generation instances are incremental; they contain
only data that has changed since the previous backup was taken. This
means that your oldest backup is a similar size to your database, but
the sizes of subsequent backups depend on the rate of change of your
data. When the oldest backup is deleted, the size of the next oldest
backup increases so that a full backup still exists.
Yet, Cloud SQL stores up to seven automated backups for each instance. This in fact allows you to restore to any specific backup, but of course, you will be delete all the data on the instance in order to restore the one in the backup.
If you are asking if it will be possible to restore only the incremental differences of an specific backup, then no, it is not possible. And it is also meant that way by the concept of Incremental Backups. You see, by definition, the incremental backup must have all the backups before. So, by restoring an "specific incremental backup into an instance" you are restoring the fullback + all the incremental backups up to the incremental backup you are requesting.
I try to configure backuping database in postgresql with pg_basebackup and WAL logs.
For now I created full backup once a week and want to backup wal logs too. But, as I understand, posgresql writes them all the time. So, how can I copy them and be shure that they are not corrupted?
Thanks
You set archive_command to a shell command that copies the WAL file to a safe archive location, so that burden is mostly on you.
When PostgreSQL runs archive_command, it assumes that the WAL file is not corrupted. Only a PostgreSQL bug or a bug in the storage system could cause a corrupted WAL segment.
There is no better protection against PostgreSQL bugs than always running the latest bugfix release, and you can invest in storage hardware that will at least detect failure.
You can also write your archive_command with a certain amount of paranoia, e.g. by comparing the md5sum of the WAL segment and its archive copy.
Another idea is to write two copies of the WAL file to different storage systems.
Current situation
So I have WAL archiving set up to an independent internal harddrive on a data logging computer running Postgres. The harddrive containing the WAL archives is filling up and I'd like to remove and archive all the WAL archive files, including the initial base backup, to external backup drives.
The directory structure is like:
D:/WALBACKUP/ which is the parent folder for all the WAL files (00000110000.CA00000004 etc)
D:/WALBACKUP/BASEBACKUP/ which holds the .tar of the initial base backup
The question I have then is:
Can I safely move literally every single WAL file except the current WAL archive file, (000000000001.CA0000.. and so on), including the base backup, and move them to another hdd. (Note that the database is live and receiving data)
cheers!
WAL archives
You can use the pg_archivecleanup command to remove WAL from an archive (not pg_xlog) that's not required by a given base backup.
In general I suggest using PgBarman or a similar tool to automate your base backups and WAL retention though. It's easier and less error prone.
pg_xlog
Never remove WAL from pg_xlog manually. If you have too much WAL then:
your wal_keep_segments setting is keeping WAL around;
you have archive_mode on and archive_command set but it isn't working correctly (check the logs);
your checkpoint_segments is ridiculously high so you're just generating too much WAL; or
you have a replication slot (see the pg_replication_slots view) that's preventing the removal of WAL.
You should fix the problem that's causing WAL to be retained. If nothing seems to have happened after changing a setting run a manual CHECKPOINT command.
If you have an offline server and need to remove WAL to start it you can use pg_archivecleanup if you must. It knows how to remove only WAL that isn't needed by the server its self ... but it might break your archive-based backups, streaming replicas, etc. So don't use it unless you must.
WAL files are incremental, so the simple answer is: You cannot throw any files out. The solution is to make a new base backup and then all previous WALs can be deleted.
The WAL files contain individual statements that modify tables so if you throw some older WALs out, then the recovery process will fail (it will not silently skip missing WAL files) because the state of the database cannot be restored reliably. You can move the WAL files to some other location without upsetting the WAL process but then you'd have to make all WAL files available again from a single location if you ever need to recover your database from some point in the past; if you are running out of disk space then that may mean recovering from some location where you have enough space to store the base backup and all WAL files. The main issue here is if you can do that fast enough to restore a full database after an incident.
Another issue is that if you cannot identify where/when a problem occurred that needs to be corrected your only option is to start with the base backup and then replay all the WAL files. This procedure is not difficult, but if you have an old base backup and many WAL files to process, this simply takes a lot of time.
The best approach for your case, in general, is to make a new base backup every x months and collect WALs with that base backup. After every new base backup you can delete the old base backup and its subsequent WALs or move them to cheap offline storage (DVD, tape, etc). In the case of a major incident you can quickly restore the database to a known correct state from the recent base backup and the relatively few WAL files collected since then.
A solution that we went for, is executing pg_basebackup every night. This would create a base backup and later on we can use pg_archivecleanup to clean up all the "old" WAL files before that base using something like
"%POSTGRES_INSTALLDIR%\bin\pg_archivecleanup" -d %WAL_backup_dir% %newestBaseFile%
Fortunately, we never had to recover yet, but it should work in theory.
In case someone found this by searching how to safely cleanup the WAL directory under a replication architecture, consider the scenario where there might be left overs from offline replicas, in this case, unused replica slots waiting for the replica to come back online and thus keeping a lot of WAL archives on the Master DB.
In our case we had an issue with a replica going down due to hardware failure, we had to recreate it along with its replica_slot on the Master DB but forgot to get rid of the previous used one. Once we cleared that out PSQL got rid of unused WALs and all was good.
You can add the script to automatically clean or remove pg_wal files. This will work in pg-11 version. If you want to use other psql version the you can simply replace the command "/usr/pgsql-11/bin/pg_archivecleanup" to /usr/pgsql-12/bin/pg_archivecleanup or 13 as per your wish.
#!/bin/bash
/usr/pgsql-11/bin/pg_controldata -D /var/lib/pgsql/11/data/ > pgwalfile.txt
/usr/pgsql-11/bin/pg_archivecleanup -d /var/lib/pgsql/11/data/pg_wal $(cat pgwalfile.txt | grep "Latest checkpoint's REDO WAL file" | awk '{print $6}')
Let's say I made a quick backup dump of my Postgres 9.3 DB through pg_dump before doing a large destructive migration and I discovered I want to undo it. No writes were performed against the DB in the meantime.
Say I run pg_restore -c -d mydb < foo.dump to load the dump back into the db. Assuming I have WAL-E set up to archive every 16mb of WAL, do I need to turn off archive_mode before performing the restore? It would not be super useful for me to archive the xlog as I'm writing the dump back into the DB, since I already have perfectly valid base backups and WAL segments archived for before the dump. Also there are serious consequences to performing xlog shipping as I'm restoring the dump, which get worse with the size of the dump.
Do you end up disabling archiving before a restore? Do you do anything else to speed things up? There's a discussion of restore performance in this post, but it doesn't cover archiving at all, unless I missed something.
You can't really turn WAL archiving on and off like that. WAL replay requires continuity.
If you turned WAL archiving off, then made changes, then turned it back on, the new WAL generated after turning it off and on again would be useless. They could not be applied to the DB, and you'll have to make a new base backup before you can resume WAL replay / PITR.
If you turn xlog shipping off during a restore, you'll want to purge your old base backup and WAL archives, then create a new base backup before resuming WAL shipping.
We're serving a Django/Postgres site running on a VM hypervisor. We're now trying to figure out our back up strategy and have two probable options:
Back up the DB directly using pg_dump
Back up the VM directly by copying the VM image
I'm with the latter as I think, I could simply back up everything that has to do with the site. I'm not sure whether I have to shut down the VM for this though.
What is a better and more recommended way of backing up a DB? Are there any reasons for not using the VM backup?
Thanks
The question basically boils down to, can you consider a hot copy of PostgreSQL's data files a backup?
The answer is: not really. PostgreSQL tries very hard through the use of WAL to ensure that its files are in a consistent state all the time and that it can survive a power failure, but starting it up from a copy of these files puts PostgreSQL into recovery mode. If the backup happened at the wrong second and PostgreSQL can't recover from the state of these files, your backup is useless. You don't want your backup/restore mechanism to depend on the recovery mechanism (unless you're dealing with "crash only" software, which PostgreSQL is not).
The probability of PostgreSQL not being able to recover from these files is not high, but it's not zero either. The probability of PostgreSQL not being able to load an SQL dump that it made, on the other hand, is zero. I prefer backup choices with lower probabilities of failure. pg_dump was designed for doing backups.
PostgreSQL recommends using pg_dump for backups, as a file system (or VM) backup requires the database to be shut down (and has other drawbacks):
http://www.postgresql.org/docs/8.1/static/backup-file.html
Edit: Also, a pg_dump backup will be significantly smaller than a filesystem dump of the same database.
There is an additional option. With PostgreSQL you can make an online backup that allows you to snapshot the file system and maintain consistency. You can see details here:
http://www.postgresql.org/docs/9.0/static/continuous-archiving.html
We use this exact method for making backups when we run PostgreSQL in a VM.