Backup Suggestions

[jimvad]

I need to add two servers (one Linux and one Windows) to my jobs list. I have reached out to the software vendor for these two VM machines for advice on how to handle their backups. I'd rather not brick the databases if you know what I mean. Does anybody have any suggestions of how I should craft the backup options for these two VMs based on the vendor's recommended backups, cautions and contingencies listed below?
Any help would be greatly appreciated.
Thanks!


Recommended Backups
The following directories on the D: drive are required to rebuild a server from scratch. Not all of the directories will exist on every outage system. If they do, though, back them up:

• dataVoice (exclude database folder)
• z_datavoice_backup1
• z_dvgispublisher_backup1
• dataVoice_special (not all utilities have this folder)
• dataVoice_bat (not all utilities have this folder)
• [utilityacronym]_data (the utility’s acronym is the name of the folder)
The best time to capture these directories is immediately after the daily backup/restart of the outage system. If you are unsure when that occurs, please contact dataVoice tech support.
Cautions and Contingencies
Be aware that older forms of full-disk backups would actually lock files as they copied. There is an active Relational Database Management System (RDBMS), specifically PostgreSQL, using directories on the D: drive. The older backup mechanisms could interfere with database operation, and earlier advice was not to do full backups of the D: drive. At a minimum those mechanisms would temporarily disable the outage system; at worst they caused database corruption.

Newer backup mechanisms, especially those taking snapshots of running virtual machines, may avoid that specific problem, but you should still be aware of a few other potential problems:

• Disk snapshots do not guaranty that the files for the PostgreSQL RDBMS have been captured in a consistent state. It is quite possible that snapshots will require some rebuilding of the database. (It is not unlike pulling the power plug on a physical server.) It is theoretically possible, however unlikely, to obtain an unrecoverable database with snapshots.
• Snapshots almost certainly affect disk I/O as they are being taken. The outage system is sensitive to degraded disk performance, especially during large outages. Please assess the disk I/O requirements before arranging snapshots.
None of this means you should not take snapshots if your standard procedure is to do that. In fact, a successful snapshot of a running VM is certainly the fastest way to recover from catastrophic loss of a server.

However, it would certainly be prudent to follow the basic backup procedures outlined above as well. It provides a contingency in the event of an unsuccessful snapshot.

[HannesK]

Hello,
and welcome to the forums.

As far as I understood, you need to make sure that the Postgres database is in a consistent state. To do that, you can just stop it before backup (that works with every application), or bring it into a consistent state. If you search the internet / forums for "postgres backup script", there will be several options / ways to do that.

Whatever option you prefer, that scripts are used as pre / post script in Veeam: https://helpcenter.veeam.com/docs/backu ... ml?ver=110

For Postgres on Linux, you could also use the Veeam Agent for Linux. It has built-in support for Postgres https://helpcenter.veeam.com/docs/backu ... ml?ver=110

Best regards,
Hannes

[jimvad]

Thanks @Hannes
Would this recommendation suffice as well:
Daily backup with AAIP (guest-processing) enabled so as to ensure that databases are preserved and in their most recent state when backups are taken.

Thanks,
Jim

[HannesK]

I'm unsure about understanding the last question: According to the software vendor, AAIP must be configured in a way that it creates a consistent state for Postgres. There are several ways to do that. Which one you take depends on your preferences.