[GE users] Open MPI tight integration in HOWTO page

Andreas.Haas at Sun.COM Andreas.Haas at Sun.COM
Thu Feb 8 13:16:29 GMT 2007


On Wed, 7 Feb 2007, Heywood, Todd wrote:

>
> Hi Andreas,
>
> I tried setting SGE_COMMLIB_DEBUG to 3 and set off the job. A qstat
> before the job hung gave a load of messages. Once the job hung, qstat
> just gave this (again):
>
> critical error: unrecoverable error - contact systems manager
> Aborted
>
> If I go to another window to do a qstat, I get normal output saying the
> job is in state "r". Doing a qdel from the second window kills the job,
> and then qstat back in the first window works normally again (with the
> extensive logging/debug messages).

Hm ... apriori I got no idea how to track into this and when 
knowledge of the deepest secrets in our commlib is needed, 
I'm not the right one to help you anyways.

Todd, if you could file an IssueZilla item for this it would 
allow our commlib guru to look after it, when he's back from 
vacation

    http://gridengine.sunsource.net/servlets/ProjectIssues

note, you must at first be registered as 'observer' before you 
can actually file issues.

> On slapd and logging: some LDAP FAQs say that logging (e.g. to
> /var/log/slapd) has a significant effect on slapd performance, and to

Hard to believe ...

> only turn logging on for debug purposes. I'm having a little discussion
> on this over on the OpenMPI list right now.

I found it, thanks.

Best regards,
Andreas

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list