Lustre discuss - May 2006 - Problem stopping lustre services

On Jul 18, 2005  11:46 -0400, Jaya Natarajan wrote:
> I am testing Lustre 1.5.1 in a Cray XD1, Linux 2.6 and having problem
> stopping lustre services. But this is a general problem I encountered in
> other OS/Lustre versions too. 
> 
> While configuring Lustre (as any component - ost, mds or a client),
> lustre services are started even if the configuration is unsuccessful
> For example if there is a problem in my configuration data, lustre
> modules are loaded,  processes started and /proc/ls/lustre directory is
> created even though the configuration is not successful. I use lconf -d,
> --cleanup and --force options to clean up lustre but this does not work.
> I can not unload the modules, /proc/fs/lustre directory still exists and
> processes ll_ost etc do not get deleted. 
You should be able to unload everything with "lconf --cleanup
--force".
If that doesn''t work, can you please file a bug with your config and
the output from "lconf -v {config}.xml" or exact steps you are doing
to get into this situation.

Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.

Hi,

I am testing Lustre 1.5.1 in a Cray XD1, Linux 2.6 and having problem
stopping lustre services. But this is a general problem I encountered in
other OS/Lustre versions too. 

While configuring Lustre (as any component - ost, mds or a client),
lustre services are started even if the configuration is unsuccessful
For example if there is a problem in my configuration data, lustre
modules are loaded,  processes started and /proc/ls/lustre directory is
created even though the configuration is not successful. I use lconf -d,
--cleanup and --force options to clean up lustre but this does not work.
I can not unload the modules, /proc/fs/lustre directory still exists and
processes ll_ost etc do not get deleted. 

I have to reboot each time to cleanup to test a modified configuration.
Is there any other way of stopping lustre services cleanly without
rebooting?

Thanks,
Jaya