[rabbitmq-discuss] RabbmitMQ consuming CPU and Memory

[Matthias Radestock]

Paul,

On 18/05/12 18:36, Paul Gross wrote:
> This server is running a build, which creates and deletes a couple of
> queues. It's possible that these creates/deletes are happening at the
> same time. Could this cause the observed behavior?

No, that shouldn't be a problem.

> The beam.smp process is definitely going nuts. Here's the output from ps
> aufx:
>
> rabbitmq 26472 0.0 0.0 4144 0 ? S May14 0:00 /bin/sh
> /usr/sbin/rabbitmq-server
> rabbitmq 26490 184 82.4 2367176 837356 ? Sl May14 9852:39 \_
> /usr/lib/erlang/erts-5.9.1/bin/beam.smp -W w -K true -A30 -P 1048576 --
> -root /usr/lib/erlang -progname erl -- -
> rabbitmq 26605 0.0 0.0 10776 0 ? Ss May14 0:00 \_ inet_gethost 4
> rabbitmq 26606 0.0 0.0 12876 36 ? S May14 0:00 \_ inet_gethost 4
>
> As you can see, beam.smp is using 184% CPU and 82.4% of memory on this
> server (4 cores, 1GB RAM).
>
> Unfortunately, by the time I was able to ssh into the server to run
> rabbitmqctl report, it said nodedown.

That's a shame. Would be really good to catch it early enough to still 
get some useful output from rabbitmqctl.

> Cluster status of node 'rabbit at bt-ci-rabbitmq' ...
> [{nodes,[{disc,['rabbit at bt-ci-rabbitmq']}]},
> {running_nodes,['rabbit at bt-ci-rabbitmq']}]

> <'rabbit at bt-ci-rabbitmq'.2.186.0>
> webhook_worker_master-testing_notifications true false
> [{"x-ha-policy","all"}]

Hmm. You have HA queues but only a single node. Is that intentional? 
While that shouldn't cause any problems, a) it is far less efficient 
than using non-HA queues in that situation, and b) single-node HA queues 
are not as well tested as their multi-node cousins.

Regards,

Matthias.