Backup UCS running on KVM guest: must I shutdown?

I am running UCS inside a KVM guest and would like to make regular backup. Do I need to shutdown UCS before taking backup? Or can I use the snapshot backup, which means the disks are frozen for a few seconds for backup without UCS knowing about it.

Thanks in advance for any hint.

Hello tony2,

KVM allows live snapshots and “normal” snapshots. In my test VMs, I have always taken live snapshots and have not encountered any problems. In a productive environment with possible users, interfaces, etc., I would always log off the users for reasons of data security, shut down the system and then take a snapshot.
Please also note the order in which UCS system roles are backed up and restored.

Regards
rheyer

Hello rheyer,

thank you for your answer. I am a bit unsure about the difference between live snapshot and “normal” snapshot. I use ProxmoxVE as the KVM host and I guess the snapshot feature in ProxmoxVE refers to live snapshot?

Regards,
Tony

Hello tony2,

i use ProxmoxVE also. In the Proxmox-Wiki you can read about the differences between the three backup modes (suspend, stop and snapshot). Maybe you get there some technical details about the snapshot functions.

Sources:

Regards
rheyer

2 Likes

Dear rheyer,

can You offer some test results, should one use STOP backup, or SUSPEND backup with UCS on Proxmox?

I use 2 directory nodes and one OX node.

I know in some cases where databases are involved on the VM, snapshot backup mode gives inconsistent backups, with Zimbra and other such servers.

Does this depend on which apps I use? What about the directory nodes? If I have 2, using stop backup should pose no problem regarding downtime, yes?

Sorry about reviving this thread, but this is the closest info about this I could find.

Thx

CadilLACi

Dear CadiLACi,

I don’t think testing is necessary here. When it comes to snapshots and databases, I know of two good practices (note the bottom line):

Option 1

  • Lock logins
  • Prevent database clients from reading/writing to the DBMS.
  • Depending on the database management system and setup (redundancy, cluster, etc.), ensure that database transactions have been processed.
  • Stop database services
  • Create snapshot

Option 2

  • Shut down VM (or database cluster)
  • Create snapshot

Does this depend on which apps I use?

Yes, definitively. Everything that generates cashing data, whether transaction-based or based on other methods, must be considered at least once.

What about the directory nodes? If I have 2, using stop backup should pose no problem regarding downtime, yes?

If both directory nodes are backed up simultaneously in snapshot stop mode, the VMs are shut down and are not available during the snapshot runtime. Source.

Sorry about reviving this thread, but this is the closest info about this I could find.

No problem, it’s better to add to a thread than to keep opening new threads for the same topic.

bottom line

It depends on availability, the systems involved, their functions, and their properties. Once you have a clear picture of this, you can create a backup and restore strategy.

Regards