Wednesday, March 19, 2014

OSG Software Release 3.2.7 - March 25th - Globus Job Manager Fix - GOC Ticket # 20220

Some sites have discovered an issue with the "globus-gram-job-manager" package distributed in OSG 3.2. The symptoms are: repeated crashes of the "globus-job-manager" process with a segmentation fault, new jobs not getting queued, and the subdirectories under "/var/lib/globus/gram_job_state" getting filled up with state files of smaller than usual size. The issue is caused by globus-job-manager not handling certain kinds of incomplete state files correctly.

The affected versions of the package are 13.53-1.2.osg32.el5 and 13.53-1.2.osg32.el6. These versions were first released in OSG 3.2.4. CEs running the 3.1 series or those that have not yet upgraded to one of these versions should not be affected.

We have a new version of the globus-gram-job-manager package available in the OSG testing repositories. The new versions are 13.53-1.3.osg32.el5 and 13.53-1.3.osg32.el6. Admins of CEs running the affected version of globus-gram-job-manager are strongly encouraged to upgrade via the following command: yum upgrade --enablerepo=osg-testing globus-gram-job-manager

OSG will be making a 3.2.7 release containing this fix on March 25th.

Please see ticket 20220 at:
https://ticket.grid.iu.edu/20220