Frequent reboots

[Mohammed Irfan PK]

Hi,
We are doing a fail over test in HP cluster environment. We reboot the machine (in fact reset) abnormally five to six times a day. This has been continuing quite a week long. Is it harmful to do such resets contineously. The fail over does not pass some scenarieos. Any recommendations?
regards
PKI

[Steven E. Protter]

Shalom,

For a specific purpose for a defined length of time, e.g. your test there will be no harm.

HP-UX servers are built to a very high standard and can endure a lot of abuse.

SEP

[freddy_21]

you need some check:
1. make sure your node will be reboot, have package running at there. command you need run is: cmviewcl

if you reboot, but no package running at the node, you cant see the package failed over.

2. you must check autorun is enable. if the status is disable, your package can't failed over to oter node. to enable using this command:
cmmodpkg -e -n -n
cmmodpkg -e

Thanks
Freddy

[Mridul Shrivastava]

This is not at all advisable to do a reset the way you are following it. It may cause to the system.

There are other ways to check the failover. like removing a netowrk cable, killing some process etc...

Refer the following threads for other ways of testing the failover:

http://forums12.itrc.hp.com/service/forums/questionanswer.do?threadId=1061805&admit=109447627+1200924169684+28353475

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=98261&admit=109447626+1200924322748+28353475

[A. Clay Stephenson]

Note that yanking a network cable or yanking a SCSI/Fibre cable does not actually test a node. In fact, your networks and disks should be so redundant that pulling any single cable or killing any single network switch does not cause a package failover. In this scenario, SG doesn't really come into play. You should also note that SG is only designed to handle a single failure so that pulling two network cables (or two disk cables) or two anythings is not a fair test. Essentially you should pull one cable and make sure that all is well and that no package failover occurs. I assume at this point that your manual package failovers (ie those done using the cmxxxx commands) work flawlessly. At this point the proper way to test for node failure is via a TOC or by yanking all the power cables.

[Mohammed Irfan PK]

Hi Clay,
Yes agreed, as part of the fail over scenarios I have carried out all the tasks like unplugging a network cable, killing the process etc. Now we have issues with Oralce RAC which locks the table during a fail over using TAF service. So after modifications in Oralce parameters, I just issue a RS (reset) from the admin. I was just wondering if frequent reboots could panic the server?

Regards
Irfan