On Mon, Oct 25, 2010 at 11:52 AM, Marek Marczykowski
<[email protected]> wrote:
> Hi,
>
> I have setup with M/S mysql resources.
> The goal is to have one IP address on MySQL master and another for slave
> (if any). When there is only (one) master - slave IP should be on the
> master.
>
> My setup produces some weird effect:
>
> When some instance of M/S MySQL fails - it is restarted
> (migration-threshold=3). It's OK for now. But when this 3 restarts
> doesn't help - whole M/S MySQL resource is stopped (also working
> instance on another node). When on node with failed instance I issue
> "crm_standby -v on", MySQL is started again (only on one, remaining node).
>
> Some details on this setup:
> 5 nodes, but only 2 permitted to run MySQL (symmetric-cluster="false").
> Failed instance is slave. resource-stickiness="100"
>
> Configuration details:
> primitive DevelDatabase ocf:heartbeat:mysql (...) \
>  meta migration-threshold="3"
> ms DevelReplDatabase DevelDatabase meta master-max="1" \
>  master-node-max="1" clone-max="2" clone-node-max="1" (...)
> primitive DevelDatabaseIP ocf:heartbeat:IPaddr2 (...)
> primitive DevelDatabaseSlaveIP ocf:heartbeat:IPaddr2 (...)
>
> colocation DevelDbIP-on-Db inf: DevelDatabaseIP DevelReplDatabase:Master
> colocation DevelDbIP2-on-Db inf: DevelDatabaseSlaveIP DevelReplDatabase
> colocation DevelDifferentDatabaseIP -1500: DevelDatabaseSlaveIP \
>  DevelDatabaseIP
>
> order Devel-ip-db-promote inf: DevelDatabaseIP:start \
>  DevelReplDatabase:promote
> order Devel-ip-db-slave inf: DevelDatabaseSlaveIP DevelReplDatabase
>
> I think the problem is colocation rules for DevelDatabaseIP, but I don't
> know why... Stopping DevelDatabaseIP resource "solves" the problem.
>
> Have anyone idea, how should I achieve correct behaviour?

Would need a hb_report to comment further.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to