Hi,
I'm having some difficulties setting up a CREAM CE (1.6.3 with Torque
2.5.4), I was wondering if anyone had similar experience or advice.
After some fiddling, I've got my CE up to the state where I can submit
jobs from a UI and have them run on a WN. I had some problems with
BUpdaterPBS segfaulting when trying to parse the output of tracelog from
this version of Torque but a one line fix to make it skip some lines it
should have been ignoring anyway got that working. I'll push this
upstream but ggus appears to be down currently.
Now though I'm having problems with BNotifier, which is refusing to
notify CREAM of changes of job state. If I restart CREAM, then it does a
manual check and correctly marks all my jobs as DONE-OK.
I did some digging and added some printfs and it seems that BNotifier
finds the new job updates in the database written by BUpdaterPBS but
rejects them on the basis that en->user_prefix == 0 at BNotifier.c:323.
From grepping the source it seems this prefix doesn't come from
BUpdaterPBS or associated tools, so it should be set by CREAM at
submission time? Clearly my job managed to resolve a user prefix because
it was correctly mapped to a user on the WN so I don't understand what
I'm doing wrong...
Does any of this ring any bells with anyone? Does anyone know what
should be setting this user prefix in the job registry?
Any advice much appreciated,
Adam
|