Update to the 11 Jan 2013 dist does not help with this issue. :(

I checked the database and have the look at the job entries that handler0 tried to stop then shutdown:

| 3088 | 2013-01-03 14:25:38 | 2013-01-03 14:27:05 |        531 | toolshed.g2.bx.psu.edu/repos/kevyin/homer/homer_findPeaks/0.1.2            | 0.1.2        | deleted_new | Job output deleted by user before job completed. | NULL         | NULL           | NULL        | NULL   | NULL   | NULL      |       1659 | drmaa://-V -j n -R y -q intel.q/ | NULL                   |              NULL |      76 |        0 | NULL            | NULL   | handler0 |      NULL |
| 3091 | 2013-01-04 10:52:19 | 2013-01-07 09:14:34 |        531 | toolshed.g2.bx.psu.edu/repos/kevyin/homer/homer_findPeaks/0.1.2            | 0.1.2        | deleted_new | Job output deleted by user before job completed. | NULL         | NULL           | NULL        | NULL   | NULL   | NULL      |       1659 | drmaa://-V -j n -R y -q intel.q/ | NULL                   |              NULL |      76 |        0 | NULL            | NULL   | handler0 |      NULL |
| 3093 | 2013-01-07 22:02:21 | 2013-01-07 22:16:27 |        531 | toolshed.g2.bx.psu.edu/repos/kevyin/homer/homer_pos2bed/1.0.0              | 1.0.0        | deleted_new | Job output deleted by user before job completed. | NULL         | NULL           | NULL        | NULL   | NULL   | NULL      |       1749 | drmaa://-V -j n -R y -q intel.q/ | NULL                   |              NULL |      76 |        0 | NULL            | NULL   | handler0 |      NULL |

So basically the job table has several of these entries what assigned to handler0 and marked as "deleted_new". When the handler0 is up, it starts stopping these jobs, after the first job has been "stopped", handler0 went crash and died. But that job was then marked as "deleted".

I think if I manually change the job state from "deleted_new" to "deleted" in the db, the handler0 will become fine. I am just concerned about how these jobs were created (like assigned to a handler but marked as "deleted_new"). 

Cheers,
D


On Mon, Jan 21, 2013 at 1:49 PM, Derrick Lin <klin938@gmail.com> wrote:
I had a close look at the code in 

galaxy-dist / lib / galaxy / jobs / handler.py
galaxy-dist / lib / galaxy / jobs / runners / drmaa.py

and found that stopping "deleted" and "deleted_new" seems to be normal routine for the job handler. Could not find any exception that caused the shutdown.

I do notice in the galaxy-dist on bitbucket, there is one commit with comment "Fix shutdown on python >= 2.6.2 by calling setDaemon when creating threads (these are still...", it seems to be relevant?

I will do the update to 11 Jan release and see if it fixes the issue.

D


On Fri, Jan 18, 2013 at 4:03 PM, Derrick Lin <klin938@gmail.com> wrote:
Hi guys,

We have updated our galaxy to 20 Dec 2012 release. Recently we found that some submitted jobs could not start (stay gray forever).

We found that it was caused by the job manager sent jobs to a handler (handler0) whose python process crashed and died.

From the handler log we found the last messages right before the crash:

galaxy.jobs.handler DEBUG 2013-01-18 15:00:34,481 Stopping job 3032:
galaxy.jobs.handler DEBUG 2013-01-18 15:00:34,481 stopping job 3032 in drmaa runner

We restarted the galaxy, handler0 is up for few seconds then died again with the same error messages except the job number moved to the next one.

We observed that the jobs it was trying to stop are all previous jobs whose status is either "deleted" or "deleted_new".

We have never seen this in the past, so wondering if there is bugs in the new release?

Cheers,
Derrick