[GE users] binary re-location

reuti reuti at staff.uni-marburg.de
Mon May 3 11:52:05 BST 2010


Am 28.04.2010 um 02:25 schrieb ron:

> Are the spool directories of the exec hosts on NFS as well??
> 
> If not, then simply stopping and restarting the qmaster won't affect the jobs.

This would imply, that also the qmaster spool directory is somewhere else like /var/spool/sge/qmaster, as otherwise also its path would change.

Depending on the necessary setup: it's also possible to bind a directory to another and it would show up at another location with `mount` or in /etc/fstab (maybe it was ment as a replacement for hard-links, which can only on Mac OS X being used for directories):

/export/opt/sge	/opt/sge	auto	bind 1 1

on the nodes to keep the names (besides using a symbolic link).

-- Reuti


> 
> -Ron
> 
> 
> --- On Wed, 4/28/10, gg3796 <gg3796 at yahoo.com> wrote:
> I believe I need to re-start execd on all nodes or reboot the cluster - (This will be the worst part)
> I can not afford killing running or pending jobs at this point.
>  
> thanks
> SB
> 
> 
> 
> 
> From: stephendennis <sdennis at univaud.com>
> To: users at gridengine.sunsource.net
> Sent: Tue, April 27, 2010 6:23:41 AM
> Subject: RE: [GE users] binary re-location
> 
> Hello List
> 
> Possibly this solution will work for you.  I have written it in
> commands to provide precision.
> 
> Assumes SGE_ROOT is /opt/sge and SGE_CELL is default, and
> that SGE the above are located on filer:/export/opt/sge/default
> 
>     cd /opt
>     mkdir sge.local
>     rsync -ruvalP --exclude default sge/ sge.local/    #< copy
>     qconf -km                #< kills the master
>     umount /opt/sge      #< maybe will have to kill local
> execd, dbwriter, other stuff before this
>     mv sge sge.save      #< move the old mountpoint out of the say (should be empty anyway)
>     mv sge.local sge      #< move the new 
>     rm -rf sge/default    #< remove the local copy of the spool, should be not there but just to be sure
>     mount filer:/export/opt/sge/default /opt/sge/default  # < mount just the cell
>     /etc/init.d/sgeqmaster start  # < start SGE
> 
> Surely there are some other details for your setup, but thats the gist of it.
> 
> Thanks
> Stephen    
> ________________________________________
> From: rayson [rayrayson at gmail.com]
> Sent: Tuesday, April 27, 2010 12:02 AM
> To: users at gridengine.sunsource.net
> Subject: Re: [GE users] binary re-location
> 
> First thing -- you will need to shutdown the qmaster during the
> directory path migration, but doing so will not affect the running and
> pending jobs.
> 
> Secondly, will you adjust the mount points so that the path names are
> the same before and after the transition?? Otherwise, you will need to
> adjust the bootstrap file:
> 
> http://gridengine.sunsource.net/nonav/source/browse/~checkout~/gridengine/doc/htmlman/htmlman5/bootstrap.html?pathrev=V62u5_TAG
> 
> Finally, you will need to adjust the "settings.[c]sh" scripts and/or
> startup scripts.
> 
> As long as you don't delete or damage the spool directory, then your
> jobs won't be lost -- which means "at most" you will cost extra
> downtime.
> 
> And I always suggest this quick and dirty hack -- if you move the
> spool directory, you can create
> a symbolic link from the original
> directory to the new one, and to SGE the path/location stays the same
> after the migration.
> 
> Rayson
> 
> P.S. I might miss a thing or 2, so test it with your test cluster
> first if you can't affort an extra minute of downtime!
> 
> 
> On 4/26/10, gg3796 <gg3796 at yahoo.com> wrote:
>> 
>> Thanks Rayson
>> sge binaries including qmaster spool directory are nfs mounted on qmaster
>> and I want to place them on local drive on qmaster.
>> 
>> thanks,
>> Syed
>> 
>> 
>> ________________________________
>> From: rayson <rayrayson at gmail.com>
>> To: users at gridengine.sunsource.net
>> Sent: Mon,
> April 26, 2010 4:47:27 PM
>> Subject: Re: [GE users] binary re-location
>> 
>> Sorry, I am not following exactly...
>> 
>> 1) SGE binaries or application binaries?
>> 
>> 2) So currently, binaries are on an NFS mounted directory, and you
>> want to place them on a local drive?
>> 
>> Rayson
>> 
>> 
>> 
>> On Mon, Apr 26, 2010 at 7:08 PM, gg3796 <gg3796 at yahoo.com> wrote:
>>> Hi:
>>> I am running 6.2u3 with binaries are nfs mounted on qmaster. I want to
>> move
>>> binaries local to qmaster. Is there a good way to do so without killing
>>> running /pending jobs?
>>> 
>>> thanks
>>> SB
>>> 
>>> 
>> 
>> ------------------------------------------------------
>> 
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=255062
>> 
>> To unsubscribe from this discussion, e-mail:
>> [users-unsubscribe at gridengine.sunsource.net].
>> 
>> 
> 
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=255081
> 
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
> 
> 
> ---------------------------------------------------------------------
> 
> 
> Notice from Univa UD Postmaster:
> 
> 
> This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged
> information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. This message has been content scanned by the Univa UD Tumbleweed MailGate.
> 
> 
> 
> ---------------------------------------------------------------------
> 
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=255130
> 
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
> 
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=255208
> 
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
>

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=255914

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list