ZFS replication DB consistency

[TSAN]

Learning ZFS replication / live migration.
ZFS replication + live migration for 2 node small biz environments a pretty neat feature - one up on ESXi local VMFS6 datastores.

I've been reading through some of the documentation regarding ZFS replication / snapshots + qemu guest agent in regards to DB application consistency.
Qumu guest agent documentation briefly touches on fs-freeze. And found some very old discussions.

Is a ZFS replication / snapshot to another host crash consistent if qemu guest agent installed? Or more config scripting is needed within qemu guest agent based a each type of different DB used?

Can anyone point me some detailed documentation / discussions regarding Proxmox / ZFS / guest agent snapshot DB consistency?

Coming from mostly ESXi with Windows guests which tie into Microsoft's VSS. Windows DBs configured for VSS to flush transaction log to DB. ESXi manual snapshotting or Veeam. But also interested in Linux VMs with sqlite / mysql / Postgres / Elasticsearch.

(Very happy Veeam onboard with Proxmox - yet to explore PBS)

 

[weehooey-bh]

Hi @TSAN

I was looking for something related to ZFS replication and VSS when I came across your post. Just in case you have not found the answer to your question...

Yes, ZFS replication triggers a freeze-thaw with VSS if you have the QEMU guest agent installed. If your database responds to VSS, then it should quiesce.

I cannot point you to the documentation because I could not find any, but I just verified in our lab that each run of the ZFS replication triggered VSS. I hope this helps.