[Samba] Do you NEED pacemaker when using Samba Clustering (ctdb)

[John Ericsson]

Thank you for your quick comprehensive reply. CTDB does the job and it does
it well. I had a bad experience on another service several years ago when a
faulty node broke the HA so I have an interest in STONITH. However, this
was not samba and when a node has failed on the samba cluster CTDB handled
it as expected.
The ipfailover quote (which is echoed on other sites) comes from here
https://documentation.suse.com/sle-ha/15-SP4/html/SLE-HA-all/cha-ha-samba.html


Again, thank you for your reply, it was appreciated.

On Sat, Nov 12, 2022 at 9:39 PM Martin Schwenke <martin at meltin.net> wrote:

> Hi John,
>
> On Sat, 12 Nov 2022 21:13:52 +0000, John Ericsson via samba
> <samba at lists.samba.org> wrote:
>
> > I have used a two-node active-active samba cluster using cstb (using
> > clustered filesystem) and it has been working fine for over 3 years.
> > I set up a test environment and configured the samba using ctdb and
> forgot
> > to turn on pacemaker. Thing is it all works fine even without pacemaker.
> > Thinking about it ... of course it does, it works as the samba team said
> it
> > works. I should have asked this question 3 years ago. .
> > So what is pacemaker brining to the party when I read stuff like
> > "The current implementation of the CTDB Resource Agent configures CTDB to
> > only manage Samba. Everything else, including IP failover, should be
> > configured with Pacemaker."
>
> I think you have answered your own question.  I am a current CTDB
> developer, but I don't know enough about the current state of Pacemaker
> to be able to say where it might do better than CTDB's IP failover.
>
> So, I guess it depends on the experience and motivations of whoever
> wrote that.  They may have encountered situations where CTDB's IP
> failover didn't meet their requirements.  If that's the case then it
> would have been nice to hear about it so we could improve CTDB.  😉️
>
> Note that we have been talking about a CTDB rewrite for quite a few
> years.  The idea would be to split CTDB into neat, logical components,
> without throwing away existing functionality.  The would be an
> improvement over CTDB's current monolithic design/implementation.  If
> we achieve this then perhaps it will be possible to replace some
> components with 3rd party components like Pacemaker. This is somewhat
> possible at the moment (obviously, since someone recommends using
> Pacemaker instead of CTDB's IP failover) but there is no way of
> completely switching off CTDB's failover code, because doing that was
> never intended.
>
> > I appreciate that there is no STONITH options.
>
> Support could be added to CTDB...
>
> peace & happiness,
> martin
>