[Linux-HA] Master-slave, stopping a slave.

c smith smith.c.tech at gmail.com
Wed Jul 8 01:54:43 MDT 2009


Dominik-

Thanks for the reply.. I'm aware that the documents advise against it, but
surely there must be a way.  I was just looking at the new DRBD 8.3.2.  It
includes a fencing handler script that, upon failure of a DRBD master, adds
a -INFINITY location constraint into the CIB that prevents the MS resource
from being 'Master' on the failed node until resync/uptodate (at which point
its deleted from the CIB)

Pasted from ./drbd-8.3.2/scripts/crm-fence-peer.sh:

new_constraint="\
<rsc_location rsc=\"$master_id\" id=\"$id_prefix-$master_id\">
  <rule role=\"Master\" score=\"-INFINITY\"
id=\"$id_prefix-rule-$master_id\">
    <expression attribute=\"$fencing_attribute\" operation=\"ne\"
value=\"$fencing_value\" id=\"$id_prefix-expr-$master_id\"/>
  </rule>
</rsc_location>"

This constraint, when in place, succesfully stops the child from running on
that node.  Wondering if it is possible to do the same with the slave
resource.

I will also try out unmanage'ing the resource and see if that will
accomplish similar. Thanks again for your input
CSMITH



> It has been stated that is it not intended to mess with the internals of
> child-management. Never to use child-instance-ids in constraints and so
> on. They will just cause you unwanted trouble.
>
> I'd probably just unmanage the drbd resource, then the cluster will not
> monitor it. Then do whatever you need to do and afterwards, let the
> cluster manage drbd again.
>
> Look into the "is-managed" meta attribute.
>
> Regards
> Dominik
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>



More information about the Linux-HA mailing list