6600...cluster node crashes ..wen rebooted..

[chindi]

Hi ,
I have a setup of a 2 - node cluster rx6600....dey both are my databse servers...
To check the HA wen v reboot one of the node i.e. wen v issue reboot command...
the Serviceguard issues a TOC and crashes...INIT...and den dumps the crash dumps...

We have sent the logs to HP..., dey say cmcld daemon is causing that TOC...
Still i would seek a clarification from you guys,,,

The serevr comes up fine wen issued
shutdown -ry 0 command.....

[Suraj K Sankari]

HI,

Need to check the syslog.log and cluster.log file ...

Please post...

Suraj

[Johnson Punniyalingam]

Possible causes:
* Heartbeat lost due to network congestion
* Network problem with cables, switch or hub.
* Software or hardware problem on node 2 causing it to crash

log entries on node1 and node2 - Needs to be Checked

Thanks,
Johnson

[chindi]

hi Suraj and Johnson,
Unfortunately i wont be able to provide you the log s at this moment , but will try to post it as soon as possible....
since dey r in cluster afetr i reboot one node , the cluster/packages comes up fine in other node...But the node wer i issue "reboot" commands it crashes....

[Johnson Punniyalingam]

>>But the node wer i issue "reboot" commands it crashes<<

In this case , None can advice at the moment we can all keep guessing only , So please wait for HP feedback and advice.

Thanks,
JOhnson

[melvyn burnard · Accepted solution]

Are you saying you issue the reboot command, as opposed to shutdown?
If so, then this is expected behaviour.
Using the reboot command is not the normal way to shut a server down. Reboot issues a kill -9 signal to processes, as opposed to shutdown which issues kill -15.

The kill -9 is a SIGKILL which when called by reboot, does a kill -9 on the running cmcld daemon, and Serviceguard sees this as a failure.
This is also why you should halt a node with the cmhaltnode command prior to installing any patch that wants to reboot the server, as the same scenario will occur.

[chindi]

Thanks Melvyn,

That was the reason i was looking for ...
thanks again ....