Re: Corosync and Pacemaker in FBSD: anyone really using this?

On Tue, Apr 30, 2024 at 05:41:37PM UTC, Alejandro Imass wrote:
> Hi all,
> 
> I've been trying to get some simple clusters to work for about a month now.
> But it all seems very unstable. I got a pgsql master/slave working but with
> the recent minor version update, new clusters aren't working anymore.
> 
> These pacemaker/corosync/crmsh ports seem more like a curiosity, and many
> things are not working as expected. I mean, I love and appreciate the fact
> that someone is trying to make this work here, but it seems very Linux
> oriented and not very stable on FBSD, much less in jails. When you ask on
> the Clusterlabs groups/lists, it seems they could care less about its use
> on FreeBSD.
> 
> Is anyone really using this in the wild ?
> 

I am using this on AWS, but I had quite a bit of hacking on my end to get
failover working.  In my case my primary issue was getting an elastic-ip
to disassociate from one instance and reassociate to the new primary.  My
use-case was to build a HA pair of HA-Proxy servers, thus the need to float
a VIP.  If I had a real physical datacenter i'd just have used CARP, but
thats not really workable in AWS or public could environments.

I'd be interested in hearing about what specific issues you are running into,
while I agree the upstream community is pretty linux-centric I was able to
it working pretty well.  Another option that might be worth exploring is
using ha-proxy itself to serve a VIP of our psql db endpoint, I've found
it to work pretty well for load balancing various TCP services and the
REST api is super handy as well.

-pete