[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

[Linux-cluster] CS5 / rgmanager pb if shutdown on 1 node


Context : cluster with two nodes (without quorum disk) :
Both nodes are running their CS5 and their services.

If on one node we launch "shutdown -h", we systematically get
the error :

<debug> Event (0:1:0) Processed
<debug> 2 events processed
clurgmgrd[29262]: <err> #47: Failed changing service status

on the other node when it tries to launch the service of first
node. However, if the first node is rebooted and the CS5 launched on it again,
then the service is well started on backup node, and it can be migrated
without any problem to the first node.

Whereas if we launch "poweroff -f" instead of "shutdown -f", it is working fine.
The difference in logs is that in the first case, the 2nd is not fencing
the 1st one because, the "shutdown -h" begins to stop the service long time
before the node is stopped, so it is normal that there is no fencing. But
the second node should have retreived the service in any case. 

Is it an already known problem ?


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]