[Pacemaker] unexpected demote request on master
Chris Feist
cfeist at redhat.com
Tue May 27 17:17:52 EDT 2014
On 05/27/14 05:38, K Mehta wrote:
> One more question.
> With crmsh, it was easy to add constraint to avoid a resource from running only
> a subset(say vsanqa11 and vsanqa12) of nodes using the following command
>
> crm configure location ms-${uuid}-nodes ms-$uuid rule -inf: \#uname ne vsanqa11
> and \#uname ne vsanqa12
> [root at vsanqa11 ~]# pcs constraint show --full
> Location Constraints:
> Resource: ms-c6933988-9e5c-419e-8fdf-744100d76ad6
> Constraint: ms-c6933988-9e5c-419e-8fdf-744100d76ad6-nodes
> Rule: score=-INFINITY
> (id:ms-c6933988-9e5c-419e-8fdf-744100d76ad6-nodes-rule)
> Expression: #uname ne vsanqa11
> (id:ms-c6933988-9e5c-419e-8fdf-744100d76ad6-nodes-expression)
> Expression: #uname ne vsanqa12
> (id:ms-c6933988-9e5c-419e-8fdf-744100d76ad6-nodes-expression-0)
> Ordering Constraints:
> Colocation Constraints:
>
> So, both expression are part of the same rule as expected.
>
>
>
> With pcs, I am not sure how to use avoid constraints if I need a resource to run
> on vsanqa11 and vsanqa12 and not on any other node.
> So I tried adding location constraint as follows:
> pcs -f $CLUSTER_CREATE_LOG constraint location vha-$uuid rule score=-INFINITY
> \#uname ne vsanqa11 and \#uname ne vsanqa12
> Even though no error is thrown, the condition after "and" is silently dropped as
> shown below
>
> [root at vsanqa11 ~]# pcs constraint show --full
> Location Constraints:
> Resource: ms-c6933988-9e5c-419e-8fdf-744100d76ad6
> Constraint: location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6
> Rule: score=-INFINITY
> (id:location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6-rule)
> Expression: #uname ne vsanqa11
> (id:location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6-rule-expr)
> Ordering Constraints:
> Colocation Constraints:
>
>
> Then I tried the following
> pcs -f $CLUSTER_CREATE_LOG constraint location vha-$uuid rule score=-INFINITY
> \#uname ne vsanqa11
> pcs -f $CLUSTER_CREATE_LOG constraint location vha-$uuid rule score=-INFINITY
> \#uname ne vsanqa12
>
> but running these two commands did not help either. Expressions were added to
> separate rules.
>
> [root at vsanqa11 ~]# pcs constraint show --full
> Location Constraints:
> Resource: ms-c6933988-9e5c-419e-8fdf-744100d76ad6
> Constraint: location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6-1
> Rule: score=-INFINITY
> (id:location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6-1-rule)
> Expression: #uname ne vsanqa12
> (id:location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6-1-rule-expr)
> Constraint: location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6
> Rule: score=-INFINITY
> (id:location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6-rule)
> Expression: #uname ne vsanqa11
> (id:location-vha-c6933988-9e5c-419e-8fdf-744100d76ad6-rule-expr)
> Ordering Constraints:
> Colocation Constraints:
>
>
> Also, tried using multistate resource name
> [root at vsanqa11 ~]# pcs constraint location
> ms-c6933988-9e5c-419e-8fdf-744100d76ad6 rule score=-INFINITY \#uname ne vsanqa11
> Error: 'ms-c6933988-9e5c-419e-8fdf-744100d76ad6' is not a resource
>
>
> Can anyone let me correct command for this ?
Which version of pcs are you using (and what distribution)? This has been fixed
upstream. (Below is a test from my system using the upstream pcs).
[root at rh7-1 pcs]# pcs constraint location D1 rule score=-INFINITY \#uname ne
vsanqa11 and \#uname ne vsanqa12
[root at rh7-1 pcs]# pcs constraint
Location Constraints:
Resource: D1
Constraint: location-D1
Rule: score=-INFINITY boolean-op=and
Expression: #uname ne vsanqa11
Expression: #uname ne vsanqa12
Thanks,
Chris
>
>
>
>
> On Tue, May 27, 2014 at 11:01 AM, Andrew Beekhof <andrew at beekhof.net
> <mailto:andrew at beekhof.net>> wrote:
>
>
> On 27 May 2014, at 2:37 pm, K Mehta <kiranmehta1981 at gmail.com
> <mailto:kiranmehta1981 at gmail.com>> wrote:
>
> > So is globally-unique=false correct in my case ?
>
> yes
>
> >
> >
> > On Tue, May 27, 2014 at 5:30 AM, Andrew Beekhof <andrew at beekhof.net
> <mailto:andrew at beekhof.net>> wrote:
> >
> > On 26 May 2014, at 9:56 pm, K Mehta <kiranmehta1981 at gmail.com
> <mailto:kiranmehta1981 at gmail.com>> wrote:
> >
> > > What I understand from "globally-unique=false" is as follows
> > > Agent handling the resource does exactly same processing on all nodes.
> For processing this resource, agent on all nodes will use exactly same
> resources (files, processes, same parameters to agent entry points, etc).
> > >
> > > In case of my resource, agent on all nodes execute same "command" to
> find score.
> > > Driver present on all nodes will make sure that the node that is to be
> promoted is the one that reports highest score as output of the "command".
> Score is reported to CM using ( /usr/sbin/crm_master -Q -l reboot -v $score)
> in monitor entry point. Until this score
> > > is reported, agent on other node will just delete the score using
> /usr/sbin/crm_master -Q -l reboot -D in monitor entry point
> > >
> > >
> > >
> > >
> > > I want to make sure that the resource does not run on nodes other than
> $node1 and $node2. To achieve this i use the following commands.
> > >
> > > pcs -f $CLUSTER_CREATE_LOG constraint location vha-${uuid}
> prefers $node1
> > > pcs -f $CLUSTER_CREATE_LOG constraint location vha-${uuid}
> prefers $node2
> > > pcs -f $CLUSTER_CREATE_LOG constraint location ms-${uuid}
> prefers $node1
> > > pcs -f $CLUSTER_CREATE_LOG constraint location ms-${uuid}
> prefers $node2
> > >
> > > Any issue here ?
> >
> > Perhaps this is not intuitive but you'd need to specify 'avoids'
> constraints for the nodes it must not run on.
> > 'prefers' only says that of all the available nodes, this one is the best.
> >
> > >
> > > Regards,
> > > Kiran
> > >
> > >
> > >
> > > On Mon, May 26, 2014 at 8:54 AM, Andrew Beekhof <andrew at beekhof.net
> <mailto:andrew at beekhof.net>> wrote:
> > >
> > > On 22 May 2014, at 11:20 pm, K Mehta <kiranmehta1981 at gmail.com
> <mailto:kiranmehta1981 at gmail.com>> wrote:
> > >
> > > > > May 13 01:38:36 vsanqa28 pengine[4310]: notice: LogActions:
> Promote vha-924bf029-93a2-41a0-adcf-f1c1a42956e5:0#011(Slave -> Master vsanqa28)
> > > > > May 13 01:38:36 vsanqa28 pengine[4310]: notice: LogActions:
> Demote vha-924bf029-93a2-41a0-adcf-f1c1a42956e5:1#011(Master -> Slave
> vsanqa27) <<<<< Why did this happen ?
> > > >
> > > > attach the file mentioned on the next line and we might be able to
> find out
> > > >
> > >
> > > Quick question, do you understand what globally-unique=false means and
> are you sure you want it?
> > > If the answer is 'yes and yes', are you sure that your agent is using
> crm_master correctly?
> > >
> > > If I run, 'tools/crm_simulate -Sx ~/Downloads/pe-input-818.bz2 -s |
> grep vha-924bf029-93a2-41a0-adcf-f1c1a42956e5', I see:
> > >
> > > vha-924bf029-93a2-41a0-adcf-f1c1a42956e5:0 promotion score on vsanqa28:
> INFINITY
> > > vha-924bf029-93a2-41a0-adcf-f1c1a42956e5:1 promotion score on vsanqa27: 2
> > >
> > >
> > > Although much of the 'INFINITY' is probably from:
> > >
> > > <rsc_location
> id="location-ms-924bf029-93a2-41a0-adcf-f1c1a42956e5-vsanqa28-INFINITY"
> node="vsanqa28" rsc="ms-924bf029-93a2-41a0-adcf-f1c1a42956e5" score="INFINITY"/>
> > >
> > > This is somewhat odd to include for a clone/master resource.
> > >
> > > _______________________________________________
> > > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> <mailto:Pacemaker at 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
> > >
> > >
> > > _______________________________________________
> > > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> <mailto:Pacemaker at 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
> >
> >
> > _______________________________________________
> > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> <mailto:Pacemaker at 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
> >
> >
> > _______________________________________________
> > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> <mailto:Pacemaker at 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
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> <mailto:Pacemaker at 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
>
>
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at 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
>
More information about the Pacemaker
mailing list