Hi All
I did not get clearly what is the impact that restarting the
job-manager-marshal would have on the running jobs
we have 400 jobs running constantly and ~ 1500 queued, and I do not
consider this number as "few jobs" from my point of view.
I would really apreciate if someone could give us a clear statement
explainig which will be affected ( only those running ? both runing and
queued ? only those not yet submitted but already landed on the CE ? )
also considering that, if this is going to be the update procedure in
"real" production ( I agree with Jeff, this is as close as we can get to
production as of today ) it is far from being optimal.
Cheers
Sergio :)
On 16, May 2008 06:15 PM, Maarten Litmaath <[log in to unmask]>
wrote:
>Hi Jeff,
>
>>>Please restart the job-manager-marshall ASAP! Better cause a few jobs
>>>to fail
>>>than wait for things to drain! CCRC'08 is a _test_.
>>
>>
>>Depends on what you want to test. If you think that updates of this
>>sort need to be rolled out during production running, then yes, you
>>should do it this way now as well. If you think during real running
>>that people will want no updates, then we should not do them now.
>
>We want to roll out a high-priority _security_ update essentially
>without paying much attention to ongoing activities at all...
>
>>It's a test, if we want it to be realistic then we should behave now
>>exactly as during production.
>
>Indeed, so this security incident even comes at the right moment!
>
>>I suspect that updates will be needed during production, so upgrading
>>now is the right thing (although one would hope that the CCRC "blessed
>>versions" table would be updated to reflect this).
>
>Good point.
Cheers
Sergio :)
---------------------------------------------
Dr. Sergio Maffioletti
Grid Group
CSCS, Swiss National Supercomputing Centre
Via Cantonale
CH-6928 Manno
Tel: +41916108218
Fax: +41916108282
email: [log in to unmask]
---------------------------------------------
|