[Pacemaker] Correct timeouts settings for UDPu

[Dmitry Koterov]

Hello.

I'm using corosync + pacemaker in UDPu mode in a geo-distributed cluster
(with large latency and some non-zero packet loss).

What are robust settings for timeouts in such situations? E.g. I suspect
that totem.token ought to be increased from 1000 to 5000 ms. Is it a good
value? Are there other timeouts/thresholds which are good to be increased
in such UDPu configuration?

[Tomcsányi, Domonkos]

Hello everyone,

I just have a quick question: I have a PostgreSQL master-slave cluster
set up using Pacemaker pretty much following this guide:
http://clusterlabs.org/wiki/PgSQL_Replicated_Cluster
I ended up with a working cluster, we tested it, failover happened just
fine, slave became master etc. so we put it in production.
Now what I see is that the pg_archive directory keeps on growing. I
thought it was supposed to get cleaned after the slave received the WAL
files needed to be in sync with the master, but that didn't happen.
I'm using the following archive command:
archive_command = 'cp %p /var/lib/postgresql/9.3/main/pg_archive/%f'

It feels weird to me, since I also saw some guides telling me to using
rsync here, or use cp, but then run a cron job to regurarly rsync the
contents of the directory to the slave node. Somehow I can't figure out
if rsync is really needed or not, to be honest I can't truly figure out
how exactly Postgres works in terms of replication.
I also tried adding the following to my RA config:
archive_cleanup_command="/usr/lib/postgresql/9.3/bin/pg_archivecleanup
/var/lib/postgresql/9.3/main/pg_archive %r"

but it didn't change the fact that my archive directory seems to be just
growing and growing, and it has some files in it that were not touched
in a month or so.

I'm confused, so I'd really appreciate any help :).

Thank you,
Domi

_______________________________________________
Pacemaker mailing list: ***@oss.clusterlabs.org
http://oss.clusterlabs.org/mailman/listinfo/pacemaker

Project Home: http://www.clusterlabs.org
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
Bugs: http://bugs.clusterlabs.org