topic Re: cluster hangs? in Operating System - HP-UX

cluster hangs?

joe_91 — Tue, 19 Nov 2002 18:49:49 GMT

Team:

When i did cmhaltnode it brought down both the packages ok. But when i do cmrunnode one package is up and running but the other is starting but both are enabled. when i try to bring down the cluster it stays as it is. I am not able to do anything with the package that says starting. Pleasee advice.

Thanks
Joe.

Re: cluster hangs?

Christopher McCray_1 — Tue, 19 Nov 2002 18:56:32 GMT

Hello,

I don't believe you are able to halt the cluster when any packages are in starting mode. How long has the package been in that state? Can you post some output? What does syslog say? the package control log?

Chris

Re: cluster hangs?

Christopher McCray_1 — Tue, 19 Nov 2002 19:13:44 GMT

Hello again,

Check out this doc:

http://www2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000063209060

Chris

test

Re: cluster hangs?

Dietmar Konermann — Tue, 19 Nov 2002 19:14:58 GMT

The package is in status "starting" as long as the package control script is running... with RUN_SCRIPT_TIMEOUT set to NO_TIMEOUT this may take forever.

You should look for running processes, especially for the shell running your control script and its children.

Have a look at it using:

UNIX95= ps -efH

In this hierarchy view you can easily determine the script's children.

Best regards...
Dietmar.

Re: cluster hangs?

Martin Johnson — Tue, 19 Nov 2002 19:42:15 GMT

Review what your packages are trying to do, and see if there are any missing dependencies, such as a file system needs to be mounted, some utility or daemon needs to be running.

It sounds like your package is waiting for something to happen and it is not happening.

HTH
Marty

Re: cluster hangs?

Jeff Schussele — Tue, 19 Nov 2002 20:00:18 GMT

Hi joe,

You need to check the log file for that pkg.
Usually named
/etc/cmcluster/pkg_name/pkg_name.log
This should give some clue as to why the start script is not completing.

Rgds,
Jeff

Re: cluster hangs?

Oleg Zieaev_1 — Thu, 21 Nov 2002 20:10:27 GMT

Joe,
you can not halt the package if it is in starting state.
Check your package.cntl file. Something is not completed in your package startup.
Special attention to application startup lines, they have to be started as daemons and release prompt for your package to continue (use & at the end).
Check the log files for package startup, i.e. your_package.cntl.log to identify on what stage package startup stall.
You can trace last program starting using pstree utility or ps -ef (sequentionaly from top to bottom).
Once you fix this, you still might consider implementing AUTO_START_TIMEOUT in cmclconf.ascii file on all your nodes.

Hope this helps,
0leg

Re: cluster hangs?

Oleg Zieaev_1 — Thu, 21 Nov 2002 20:20:25 GMT

In the package.asci file we have :
RUN_SCRIPT /etc/cmcluster/usg_pkg/usage.cntl
RUN_SCRIPT_TIMEOUT 1800
HALT_SCRIPT /etc/cmcluster/usg_pkg/usage.cntl
HALT_SCRIPT_TIMEOUT 1800

This are the timeouts you have to change.

Sorry for confusion.
Hope this helps.
0leg