[GE users] Migrating sgemasterd to another node on RHEL5

Andy Schwierskott andy.schwierskott at sun.com
Wed Sep 26 11:39:23 BST 2007


Hi,

the HA agent for SGE of Sun Cluster (Sun's HA solution) does not use the
"sgemaster -migrate" script.

You could look in the source code of the sge_shadowd what needs to be done
when a new master should be started. It's in principle quite simple:

    - check (as much as possible) if the old qmaster is really down and *no*
      "lock" file is created in the qmaster spool directory (this file is
      created on proper shutdown). Also check if the shceduler is down (and
      kill it if necessary, "kill -9" is fully ok for the scheduler in this
      case

    - once you sure that no qmaster is running start a new qmaster. This
      could be on the same host or on a different host.

Make sure the new qmaster host is an admin host.

The new qmaster while write the "act_qmaster" file - the execd's will read
within one load report interval this file and send their load reports to the
new qmaster host. All clients will automatically contact the new qmaster
hosts when they are started (and qmon will re-read the "act_qmaster" file
when it fails to contact qmaster).

Andy


On Wed, 26 Sep 2007, Andreas.Haas at Sun.COM wrote:

> On Mon, 24 Sep 2007, Vincent Bernat wrote:
>
>> 
>> Hi !
>> 
>> Instead of using sge_shadowd process to ensure that sge_master is always
>> available, I would like to use rgmanager (or clurgmgrd) which comes with
>> RHEL 5. The spooldb will be on a GFS volume and then will be available to
>> the target node if the current node becomes unavailable.
>> 
>> Does someone has already setup such an environment? Apart from launching
>> sge_masterd on the new node, what file should I modify to make other nodes
>> aware of this change? Should I use the -migrating option? Do I need to
>> setup a virtual IP?
>
> I don't know anythig about the rgmanager, but when you run
>
>   # sgemaster -migrate
>
> on the new host it shuts down old qmaster, changes all files that are to be 
> changed and launches the new master.
>
> Regards,
> Andreas
>
> http://gridengine.info/
>
> Sitz der Gesellschaft: Sun Microsystems GmbH, Sonnenallee 1, D-85551 
> Kirchheim-Heimstetten
> Amtsgericht Muenchen: HRB 161028
> Geschaeftsfuehrer: Marcel Schneider, Wolfgang Engels, Dr. Roland Boemer
> Vorsitzender des Aufsichtsrates: Martin Haering
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net
>
>

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