On 05/25/2011 11:03 AM, Steffen Neumann wrote:
> On Wed, 2011-05-25 at 09:30 -0400, John Young wrote:
> ...
>> I can manually start sgeexecd and it comes up
> ...
>> I have looked around for some log that might give me a clue what is
>> happening, but so far I have not found anything.
> We had something like that on Ubuntu,
> and it was because the execd was started 
> before the network was up. We had a log in /tmp 
> 
> Yours,
> Steffen

Thanks -- I took a look in /tmp and there was a logfile named
execd_messages.2492.  It says:

[root@compute-1-31 tmp]# cat execd_messages.2492
05/25/2011 11:34:06|  main|compute-1-31|C|can't create directory 
"compute-1-31": No such file or
directory

Sure enough -- when I look in /opt/gridengine/default/spool
there is no directory "compute-1-31", but as soon as I
manually start sgeexecd, it is created.

I am guessing that this means that there is some sort of race
condition between gridengine starting and all of the filesystem
mounts being finished.  (My /opt/gridengine directory on the
client is on a local disk, not NFS mounted, so I would think
that NFS starting should not be an issue...)

I think Reuti mentioned in another email that he has seen this
problem with openSUSE and solved it by renaming S50sgeexecd to
S99sgeexecd.  That would likely work for subsequent reboots,
but does anyone know how I could force Rocks to install it that
way to begin with??

Thanks for the help!

JY
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to