[Pacemaker] using clone score in location

[Andrew Beekhof]

On Tue, Sep 1, 2009 at 2:40 AM, Geoff
Lovett<geoff.lovett at lightningsource.com> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Tried "colocation proxy-colo inf: proxy_vip proxy_group"

well thats not going to help, you need to use the _clone_, eg.

   colocation proxy-colo inf: proxy_vip proxy_pingd_clone

> instead of
> colocation, failed squid, but proxy_vip still says put.
>
> I came up with a solution, but it's not optimal.  I added
> on-fail="standby" to squid's monitor, and added clone-node-max="1" to
> proxy_group_clone.  This puts the whole node in standby node if squid
> fails, causing proxy_vip to move.  Clone-node-max=1 was necessary to
> keep the failed node's clone from restarting on the healthy one.
>
> If I had unrelated resources managed by pacemaker on these nodes, this
> would fail those things over also.  Any additional ideas that could
> solve this scenario would be appreciated.
>
> Thanks!
> Geoff
>
> Andrew Beekhof wrote:
>> On Sat, Aug 29, 2009 at 12:21 AM, Geoff
>> Lovett<geoff.lovett at lightningsource.com> wrote:
>> Hello all, I'm trying to use a clone's score in a location constraint,
>> and it's not going well.  Basically, I've got an IP I want to run on a
>> node with any healthy clone.  My clone is of a group.  If any member of
>> the group fails to run, I'd like the IP to migrate to a healthy node.
>> Currently, after I cause a group member to fail, then the IP does nothing.
>>
>> Here's output from crm configure show.  I've tried many different
>> location commands besides this one:
>>
>>> Thats your problem, you need a colocation constraint. Not a location one.
>>
>> primitive dansguardian lsb:dansguardian \
>>        op monitor interval="5s"
>> primitive proxy_pingd ocf:pacemaker:pingd \
>>        params host_list="1.2.3.4" multiplier="1000" dampen="5s" \
>>        op monitor interval="15s" timeout="5s"
>> primitive proxy_vip ocf:heartbeat:IPaddr \
>>        params ip="1.2.3.5" \
>>        op monitor interval="5s" \
>>        meta resource_stickiness="500"
>> primitive squid lsb:squid \
>>        op monitor interval="5s" \
>>        op stop interval="0" timeout="180s"
>> group proxy_group squid dansguardian
>> clone proxy_group_clone proxy_group \
>>        meta globally-unique="false"
>> clone proxy_pingd_clone proxy_pingd \
>>        meta globally-unique="false"
>> location proxy_location proxy_vip \
>>        rule $id="proxy_location-rule" -inf: proxy_group_clone lt 0
>>
>> ptest -L -s shows score 0 for proxy_group_clone when a node is healthy,
>> and -1000000 when one a member of proxy_group has failed.
>>
>> I may be going about this wrongly... however I don't wish to add the IP
>> to the group, as one of the group members (squid) takes a long time to
>> stop, and I don't want the IP to be stuck on a non-functioning node
>> during this time.
>>
>> Currently using Pacemaker 1.0.4 / Heartbeat 2.99.  Any ideas would be
>> appreciated.
>>
>> Thanks,
>> Geoff
>>>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>
>
>> _______________________________________________
>> Pacemaker mailing list
>> Pacemaker at oss.clusterlabs.org
>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
>
> - --
>
> Geoff Lovett
> Sr. Network Engineer
>
> Lightning Source Inc.
> 1246 Heil Quaker Blvd.
> La Vergne, TN 37086
> geoff.lovett at lightningsource.com
> www.lightningsource.com
> Phone: 615-213-4610
> Fax: 615-213-4727
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.9 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iEYEARECAAYFAkqcbWoACgkQ8nT/kBYXqgwr9QCfWqM6Y/i108l2IyENMflWc7az
> OhgAnAhSh2/+RAxzhJaVcDZtFnbcM+wc
> =oDZk
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>