[GE users] strange OS X behaviour with 6.0u1 built from CVS 6.0u1_TAG

Andy Schwierskott andy.schwierskott at sun.com
Fri Sep 24 15:02:09 BST 2004


Chris,

(I assume it's 6.0u1)

1. Does it happen with classic spooling as well?

2. Can you check if the config gets really spooled to BDB:

      $SGE_ROOT/utilbin/sol-sparc64/spooledit list SUBMITHOST

    and

    $SGE_ROOT/utilbin/sol-sparc64/spooledit dump SUBMITHOST:<hostname>


Andy

> {summary: SGE "forgets" about configured submit hosts on OS X}
>
>
> Has anyone seen symptoms like this before? Should I file an issue?
>
>
> 1- after installation things work fine
>
> 2- SGE is stopped, then restarted
>
> 3- 'qrsh' fails to work because 'this host is not a submit host'
>
> 4- "qconf -ss" shows that, yes, the host is a submit host
>
> 5- qrsh still does not work ('not a submit host')
>
> 6- "qconf -as" is used to re-add the host as a submit host
>
> 7- output of "qconf -ss" shows submit host listed *twice*
>
> 8- qrsh works again
>
> 9- SGE stopped/started
>
> 10- qrsh no longer works
>
>
> When I first saw this on a large Apple cluster I was willing to blame our 
> experimental config where we were using BerkeleyDB spooling to an Apple XSAN 
> filesytem shared between 4x failover shadow masters. (works great BTW -- a 
> good way to get a HA Apple SGE cluster...)
>
> But now we are seeing this behaviour on our Apple powerbooks with vanilla 
> configurations:
>
>> bvtibook:~ root# SystemStarter stop SGE
>> Welcome to Macintosh.
>> Shutting down Grid Engine execution daemon
>> Shutting down Grid Engine scheduler
>> Shutting down Grid Engine qmaster
>> Startup complete.
>> Hangup
>> bvtibook:~ root# SystemStarter start SGE
>> Welcome to Macintosh.
>> Initializing network
>> Starting SGE Cluster
>> starting sge_qmaster
>> starting sge_schedd
>> starting sge_execd
>> Startup complete.
>> Hangup
>> bvtibook:~ root# qconf -ss
>> bvtibook.cluster.private
>> bvtibook:~ root# qrsh hostname
>> error: denied: host "bvtibook.cluster.private" is no submit host
>> bvtibook:~ root# qconf -as bvtibook.cluster.private
>> bvtibook.cluster.private added to submit host list
>> bvtibook:~ root# qrsh hostname
>> bvtibook
>> bvtibook:~ root# qconf -ss
>> bvtibook.cluster.private
>> bvtibook.cluster.private
>
>
> This is the most interesting bit:
>
>> bvtibook:~ root# qconf -ss
>> bvtibook.cluster.private
>> bvtibook.cluster.private

---------------------------------------------------------------------
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