Applications • Re: Backup with 'application aware processing' turn off Postgres & Oralce processing

Thanks for your interest in this issue.

This is the general arrangement in mirror:
N1 - Primary [Active node]
N2 - Backup [Passive node]
Arbiter [Witness server]

N1 sends "I'm alive" messages to N2 and arbiter
N2 sends "I'm alive" messages to N1 and arbiter

N1, N2, Arbiter are VMWare VM
When Veeam backs up N1,
runs pre-freeze script
VMWare snapshot
runs post-thaw script

The pre-freeze script is supplied by vendor.
The pre-freeze script when run flushes all memory buffers to disk.
The pre-freeze script when run stops sending "I'm alive" messages to N2 and arbiter

The post-thaw script is supplied by vendor.
The post-thaw script when run re-starts sending "I'm alive" messages to N2 and arbiter

If there is a large difference in time between the run of the post-thaw script and pre-freeze script
then N2 assumes that N1 has gone down and attempts to take over control. The time that N2 waits is set by
the QOS value.

When post-thaw script gets run on N1 it tries to take control again.
Chaos!

On checking the Veeam logs there is quite a bit of processing Oracle and Postgres between
the run of pre-freeze script and post-thaw script

Statistics: Posted by phillip_from_oz — Nov 13, 2024 2:37 am

---