How to isolate cause of system crash

[frederick hannah]

I have an rp3440-4, hpux 11.11 that rebooted itself as cisco switch work was being done in the same computer room. I have checked syslog and dmesg for clues, but I cant find why the server rebooted. It wasnt a power issue as the switch work was being done on the opposite end of the room and only the rp3440 was impacted. Any ideas of where I can look to find the cause of the outage?

[Pete Randall]

Since dmesg is going to be current rather than historical that won't do much good. Anything under /var/adm/crash? You say you checked syslog - did you check the old syslog? /var/adm/syslog/OLDsyslog.log


Pete

[frederick hannah]

Yes, I checked syslog and OLDsyslog. Nothing at all. This is the second time that work on that switch has caused this server to reboot. But only this one server.

[Steven E. Protter]

Shalom,

If the system is part of a serviceguard cluster, switch work could trigger a TOC, transfer of control and crash by interrupting the heartbeat.

To get good help, you are going to need to provide a lot more detail on your environment.

SEP

[frederick hannah]

The server is avi package-less cluster. Service guard is only there for lan failover.
However, the cluster was down after the reboot so there is a chance MCSG may have had a hand in it. I restarted the cluster, the response was cluster not started, but the cluster did restart w/o incident. Strangely enough, there were no MCSG-related messages referencing cluster activity in syslog.

[Raj D.]

Frederick,

- Check GSP chasis logs.
- Check /var/adm/crash/crash.x # where x is crash number.
- check /var/tombstones/ for any file with grep cpu # if true cpu caused the crash.
- check /etc/shutdownlog
- If it is due to the power trip, there will not be any log in the server. But GSP log will have power failure log.
- Check the timestamp of last portion of /var/adm/syslog/OLDsyslog.log and first portion of /var/adm/syslog/syslog.log and you can figure out what time it was rebooted and if it is related to the maintenance.


Cheers,
Raj.

[S.N.S]

Hi,

The MP in IA 64 or GSP in PA-RISC would almost certainly have some related entry...
That would be a start to get to the root cause...

Let us know your progress, findings.

HTH
SNS