[Pacemaker] It affects it that the update of the attribute by attrd is late, and a resource starts with a standby node.

renayama19661014 at ybb.ne.jp renayama19661014 at ybb.ne.jp
Mon Nov 29 02:18:47 UTC 2010


Hi,

We constituted a cluster by two node constitution.
It is constitution complicated slightly that included two pingd in
constitution.

We confirmed a phenomenon in the next procedure.

Step1) 192.168.40.3 addresses invalidate the understanding of ping.
Step2) Start two nodes and send trac1383.crm.

============
Last updated: Mon Nov 29 10:35:09 2010
Stack: Heartbeat
Current DC: srv02 (27456b6d-bb8e-445b-9baf-725a6b9417c6) - partition with
quorum
Version: 1.0.10-b2e39d318fda501e2fcf223c2d039b721f3679a9
2 Nodes configured, unknown expected votes
12 Resources configured.
============

Online: [ srv01 srv02 ]

 Resource Group: TESTgroup1
     TESTIPaddr (ocf::heartbeat:IPaddr2):       Started srv02
     TESTDummy01        (ocf::pacemaker:Dummy): Started srv02
     TESTDummy02        (ocf::pacemaker:Dummy): Started srv02
     TESTDummy03        (ocf::pacemaker:Dummy): Started srv02
 Resource Group: groupStonith1
     prmStonithN1-1     (stonith:external/stonith-helper):      Started srv02
     prmStonithN1-2     (stonith:external/ssh): Started srv02
     prmStonithN1-3     (stonith:meatware):     Started srv02
 Resource Group: groupStonith2
     prmStonithN2-1     (stonith:external/stonith-helper):      Started srv01
     prmStonithN2-2     (stonith:external/ssh): Started srv01
     prmStonithN2-3     (stonith:meatware):     Started srv01
 Clone Set: clnPingd1
     Started: [ srv01 srv02 ]
 Clone Set: clnPingd2
     Started: [ srv01 srv02 ]
 Clone Set: clnTESTmssd
     Started: [ srv01 srv02 ]
 Clone Set: clnTESTopsc
     Started: [ srv01 srv02 ]
 Clone Set: clncrond
     Started: [ srv01 srv02 ]
 Clone Set: clnportmap
     Started: [ srv01 srv02 ]
 Clone Set: clnsnmpd
     Started: [ srv01 srv02 ]
 Clone Set: clnvsftpd
     Started: [ srv01 srv02 ]
 Clone Set: clnxinetd
     Started: [ srv01 srv02 ]


We hoped that a TESTgroup1 resource started in a node of active node(srv01),
but the result was started in the node of the standby(srv02).

The problem seems to be caused by the fact that the attribute change of the
active node of attrd is late somehow or other.

Can we evade this problem by setting?

 * hb_report attached it to Bugzilla.
  * http://developerbugs.linux-foundation.org/show_bug.cgi?id=2528

Best Regards,
Hideo Yamauchi.




More information about the Pacemaker mailing list