[Pacemaker] [Problem] Order which combined a master with clone is invalid.

[renayama19661014 at ybb.ne.jp]

Hi David,

Thank you for comments.

> http://www.clusterlabs.org/doc/en-US/Pacemaker/1.1/html/Pacemaker_Explained/ch06s03s02.html

I confirmed it in INFINITY.

(snip)
      <rsc_colocation id="rsc_colocation-1" rsc="msDrPostgreSQLDB" rsc-role="Master" score="INFINITY" with-rsc="clnPingd"/>
      <rsc_order first="clnPingd" id="rsc_order-1" score="INFINITY" symmetrical="false" then="msDrPostgreSQLDB"/>
(snip)


By the start only for Master nodes, it was controlled well.

[root at drbd1 ~]# crm_mon -1 -f 
============
Last updated: Mon Jul 23 08:24:38 2012
Stack: Heartbeat
Current DC: NONE
1 Nodes configured, unknown expected votes
2 Resources configured.
============

Online: [ drbd1 ]


Migration summary:
* Node drbd1: 
   prmPingd:0: migration-threshold=1 fail-count=1000000

Failed actions:
    prmPingd:0_start_0 (node=drbd1, call=4, rc=1, status=complete): unknown error


However, the problem occurs when I send cib after the Slave node started together.

============
Last updated: Mon Jul 23 08:35:41 2012
Stack: Heartbeat
Current DC: drbd2 (6d4b04de-12c0-499a-b388-febba50eaec2) - partition with quorum
Version: 1.0.12-unknown
2 Nodes configured, unknown expected votes
2 Resources configured.
============

Online: [ drbd1 drbd2 ]

 Master/Slave Set: msDrPostgreSQLDB
     Masters: [ drbd2 ]
     Slaves: [ drbd1 ] -------------------------------> Started and Status Slave.
 Clone Set: clnPingd
     Started: [ drbd2 ]
     Stopped: [ prmPingd:0 ]

Migration summary:
* Node drbd1: 
   prmPingd:0: migration-threshold=1 fail-count=1000000
* Node drbd2: 

Failed actions:
    prmPingd:0_start_0 (node=drbd1, call=4, rc=1, status=complete): unknown error

Best Regards,
Hideo Yamauchi.



--- On Sat, 2012/7/21, David Vossel <dvossel at redhat.com> wrote:

> 
> 
> ----- Original Message -----
> > From: renayama19661014 at ybb.ne.jp
> > To: "PaceMaker-ML" <pacemaker at oss.clusterlabs.org>
> > Sent: Friday, July 20, 2012 1:39:51 AM
> > Subject: [Pacemaker] [Problem] Order which combined a master with clone is    invalid.
> > 
> > Hi All,
> > 
> > We confirmed movement of order which combined a master with clone.
> > We performed it by a very simple combination.
> > 
> > Step1) We change it to produce start error in Dummy resource.
> > 
> > (snip)
> > dummy_start() {
> > return $OCF_ERR_GENERIC
> >     dummy_monitor
> > (snip)
> > 
> > Step2) We start one node and send cib.
> > 
> > 
> > However, as for the master, it is done start even if start of clone
> > fails.
> > And it becomes the Slave state.
> 
> Not a bug,  You are using advisory ordering in your order constraint.
> 
> http://www.clusterlabs.org/doc/en-US/Pacemaker/1.1/html/Pacemaker_Explained/ch06s03s02.html
> 
> > 
> > ============
> > Last updated: Fri Jul 20 15:36:10 2012
> > Stack: Heartbeat
> > Current DC: NONE
> > 1 Nodes configured, unknown expected votes
> > 2 Resources configured.
> > ============
> > 
> > Online: [ drbd1 ]
> > 
> >  Master/Slave Set: msDrPostgreSQLDB
> >      Slaves: [ drbd1 ]
> >      Stopped: [ prmDrPostgreSQLDB:1 ]
> > 
> > Migration summary:
> > * Node drbd1:
> >    prmPingd:0: migration-threshold=1 fail-count=1000000
> > 
> > Failed actions:
> >     prmPingd:0_start_0 (node=drbd1, call=4, rc=1, status=complete):
> >     unknown error
> > 
> > 
> > We confirmed it just to make sure in Pacemaker1.1.7.
> > However, the problem was the same.
> > 
> > ============
> > Last updated: Fri Jul 20 22:53:22 2012
> > Last change: Fri Jul 20 22:53:09 2012 via cibadmin on fedora17-1
> > Stack: corosync
> > Current DC: fedora17-1 (1) - partition with quorum
> > Version: 1.1.7-e6922a70f742d3eab63d7e22f3ea0408b54b5dae
> > 1 Nodes configured, unknown expected votes
> > 4 Resources configured.
> > ============
> > 
> > Online: [ fedora17-1 ]
> > 
> >  Master/Slave Set: msDrPostgreSQLDB [prmDrPostgreSQLDB]
> >      Slaves: [ fedora17-1 ]
> >      Stopped: [ prmDrPostgreSQLDB:1 ]
> > 
> > Migration summary:
> > * Node fedora17-1:
> >    prmPingd:0: migration-threshold=1 fail-count=1000000
> > 
> > Failed actions:
> >     prmPingd:0_start_0 (node=fedora17-1, call=14, rc=1,
> >     status=complete): unknown error
> > 
> > 
> > I think that this problem is similar to the bug that I reported
> > before.
> > 
> >  * http://bugs.clusterlabs.org/show_bug.cgi?id=5075.
> > 
> > Is this problem a bug?
> > Or can we be improved by setting?
> 
> see advisory ordering
> 
> > 
> > Best Regards,
> > Hideo Yamauchi.
> > 
> > 
> > _______________________________________________
> > 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
> > 
>