On 7/10/23 08:19, Felix Huettner via dev wrote:
> previously it was not possible to set the probe interval for the
> connection from a relay to the backing ovsdb-server. With this change it
> is now possible using the
> `ovsdb-server/set-active-ovsdb-server-probe-interval` command.
> 
> The command `ovsdb-server/set-active-ovsdb-server-probe-interval` is
> already used to set the probe interval for active-backup replication.
> However this is mutally exclusive with being a relay and the case for
> using the command is the same. Therefor we decided to reuse it instead
> of adding a new one.

Hi, Felix.  Thanks for the patch!

It is important to be able to set the probe interval for this direction,
I agree.  And in the absence of the unified configuration method, the
appctl seems to be as an acceptable solution.

Though I don't think we should mix configuration for active-backup with
relays.  The 'active-ovsdb-server' part in the name is a direct reference
to an active-backup replication.  Terminology is a bit different:

  AB service model    relay service model
  ----------------    -------------------
  backup              relay
  active              relay source

So, maybe something like 'set-relay-source-probe-interval'?

And I'm not sure it's mutually exclusive to have active-backup and a
relay in the same process.  What prevents it?


For the other part of the change, it's not mentioned in the commit message,
but mentioned in the NEWS file that the default values is changed to 60s.
I'm not sure that's a right thing to do.  In the active-backup model, the
usual use-case is to back up a standalone database.  And it's important to
keep the connection alive for long since there is no other source to
replicate anyway.  On the other hand, relays are most likely used in pair
with a clustered service model, and supposed to be faster in reaction to
cluster changes.  There are likely other servers that we can re-connect to.

I'd keep the default inactivity probe value for relays as is, as it might
be important for smaller clusters to fail-over faster.  Users who frequently
see disconnections due to inactivity probes on relay connection may evaluate
reasons and adjust the value for themselves, if necessary.

What do you think?

Best regards, Ilya Maximets.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to