[torquedev] clearing exec_host on job requeue

Åke Sandgren ake.sandgren at hpc2n.umu.se
Wed Feb 14 23:46:05 MST 2007


On Wed, 2007-02-14 at 18:52 -0700, Garrick Staples wrote:
> CRI has a trouble ticket open about a job's exec_host not being cleared
> when it is requeued.  Apperently this annoys some sysadmins and breaks
> some 3rd party things like clumon.
> 
> I think I just found a bug that pre-dates TORQUE and is fixed with a
> single character patch!  I need others to look at this and tell me I'm
> not crazy.
> 
> I've already committed it to trunk, but this is trivial for 2.1 as well.
> 
> $ svn diff -r1242:1243 src/server/req_jobobit.c 
> Index: src/server/req_jobobit.c
> ===================================================================
> --- src/server/req_jobobit.c    (revision 1242)
> +++ src/server/req_jobobit.c    (revision 1243)
> @@ -1419,7 +1419,7 @@
>  
>        /* Now re-queue the job */
>  
> -      if ((pjob->ji_qs.ji_svrflags | JOB_SVFLG_HOTSTART) == 0) 
> +      if ((pjob->ji_qs.ji_svrflags & JOB_SVFLG_HOTSTART) == 0) 
>          {
>          /* in case of server shutdown, don't clear exec_host */
>          /* will use it on hotstart when next comes up       */

At the very least the "|" construct will always result in the == 0 being
false.

And yes the fix looks very good.
Nice catch.

-- 
Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
Internet: ake at hpc2n.umu.se   Phone: +46 90 7866134 Fax: +46 90 7866126
Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se



More information about the torquedev mailing list