[ClusterLabs] clutering rabbitmq

[Oyvind Albrigtsen]

There's also the rabbitmq-cluster agent, but that is also made for
OSP, and seems to wipe data during the start-action.

So the best way is probably to run it via the systemd unit file. Run
"pcs resource list systemd" to find it, and create it the same way
(it doesnt take any parameters, so settings has to be set in e.g.
the config files in /etc/ or similar).


Oyvind Albrigtsen

On 25/06/24 11:48 GMT, Damiano Giuliani wrote:
>I Ken, thanks for answering.
>Yes unfortunately the rabbitmq-cluster agent wipe everything and losing our
>quorum queue is not an option 😄 we just learnt after a power interrupion
>this! I was super confident about the agent really didn't expected a
>complete wipe and rebuild!lession learned!
>
>If somebody con join the conversation and share how achieve high
>availability rabbimq with pacemaker helping me I would really appreciate!
>
>On Mon, Jun 24, 2024, 5:19 PM Ken Gaillot <kgaillot at redhat.com> wrote:
>
>> Hi,
>>
>> The rabbitmq-cluster agent was written specifically for the OpenStack
>> use case, which is fine with recreating the cluster from scratch after
>> problems. I'm not sure about the other two, and I'm not really familiar
>> with any of the agents. Hopefully someone with more experience with
>> RabbitMQ can jump in.
>>
>> On Thu, 2024-06-20 at 10:33 +0200, Damiano Giuliani wrote:
>> > Hi,
>> >
>> > hope you guys can help me,
>> >
>> > we have builded up a rabbitmq cluster using pacemaker resource called
>> > rabbitmq-cluster.
>> > everything works as exptected till for maintenance reason, we shutted
>> > down the entire cluster gracefully.
>> > at the startup we noticed all the user and permissions were dropped
>> > and probably also the quorum queues.
>> > So investigating the resource agent (rabbitmq-cluster), i find out it
>> > callss this wipe function
>> >
>> > rmq_wipe_data()
>> > {
>> > rm -rf $RMQ_DATA_DIR > /dev/null 2>&1
>> > }
>> >
>> > when the first start function is called
>> >
>> > rmq_start_first()
>> > {
>> > local rc
>> >
>> > ocf_log info "Bootstrapping rabbitmq cluster"
>> > rmq_wipe_data
>> > rmq_init_and_wait
>> > rc=$?
>> >
>> > So probably when all the cluster is fired up by pacemaker all the
>> > rabbitmq istances are wiped out.
>> >
>> > the rabbitmq-cluster is quite old (3-4yo) and probably didnt take
>> > into account quorum queues which are presistent, so a full wipe is
>> > not acceptable.
>> >
>> > So i moved to the RA called rabbitmq-server-ha which is quite huge
>> > and big script but im a bit lost because i notice also this one seems
>> > cleans mnesia folder.
>> >
>> > So the third and last one is the RA rabbitmq-server  which seems
>> > simple resoruce but not manage cluster status but only simple actions
>> > like start stop etc.
>> > i could probably build the cluster using this one + rabbitmq.conf
>> > file where i defined cluster istances, something like this.
>> >
>> https://www.rabbitmq.com/docs/cluster-formation#peer-discovery-classic-config
>> >
>> > so im a bit lost because seems there is no easy way to build up a
>> > rabbitmq cluster using pacemaker.
>> >
>> > can you guys help me heading on the correct way?
>> >
>> > thanks
>> >
>> > Damiano
>> >
>> >
>> >
>> >
>> > _______________________________________________
>> > Manage your subscription:
>> > https://lists.clusterlabs.org/mailman/listinfo/users
>> >
>> > ClusterLabs home: https://www.clusterlabs.org/
>> --
>> Ken Gaillot <kgaillot at redhat.com>
>>
>> _______________________________________________
>> Manage your subscription:
>> https://lists.clusterlabs.org/mailman/listinfo/users
>>
>> ClusterLabs home: https://www.clusterlabs.org/
>>

>_______________________________________________
>Manage your subscription:
>https://lists.clusterlabs.org/mailman/listinfo/users
>
>ClusterLabs home: https://www.clusterlabs.org/