[torqueusers] max job id number?

Glen Beane glen.beane at gmail.com
Wed Apr 2 05:27:35 MDT 2008


On Wed, Apr 2, 2008 at 7:02 AM, Åke Sandgren <ake.sandgren at hpc2n.umu.se>
wrote:

> On Wed, 2008-04-02 at 12:52 +0200, Bas van der Vlies wrote:
> > Åke Sandgren wrote:
> > > On Wed, 2008-04-02 at 00:05 -0400, Brock Palen wrote:
> > >> I know someone must have asked this before, but we just rolled over 1
> > >> million jobs, and already have over 1.3 million in gold.
> > >>
> > >> What is the max job id?  Is it just 32 bit int?  (2^32)/2-1? Just
> > >> curios, as we do many tools based on job ids always getting bigger.
> > >
> > > Yes, interesting problem. We are at 1.4M jobs and counting...
> > > Fortunately that cluster is going out of service in a few month
> > > (hopefully)
> > >
> > > Anyone who knows?
> > >
> > source torque 2.3.0
> >
> > include/server.h:
> >   - int  sv_jobidnumber;    /* next number to use in new jobid  */
>
>
> Then I would like to see this get changed into an unsigned long.
> Clusters are getting larger and jobs will flow through faster...



That wouldn't be difficult, but it would require a change to the job struct
(the char array that stores the full job ID would have to be longer to
accommodate longer job id numbers). We've already had to do this a couple
times, and even though we have code that will upgrade .JB files when we
change things like this, some users have found that it doesn't always work
well to upgrade a running cluster when a .JB file upgrade is required - the
safest thing to do is drain the system of running jobs and then do the
upgrade.

Since we've had minor job struct changes in 2.2 and 2.3 I'd like to try to
hold out for a little while
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20080402/20b3a3da/attachment-0001.html


More information about the torqueusers mailing list