[Mauiusers] Maui resource reservations problems
lapan_mv at inbox.ru
Mon Nov 21 11:04:03 MST 2005
I'm new to maui/pbs world, so please sorry if my question is obvious :-).
We are using Torque/Maui to perform resource management on our
cluster. Torque is 1.2.0p6, maui is 3.2.6p13.
Everything works fine except one 'feature': sometimes, maui segfaulted
in the middle of work.
I've noticed strange correlation between this behaviour and reservations
settings of scheduler. For example:
We've set one user's reservation for a bunch of nodes:
# setres -u lapan_mv n[0-1][0-9]
User 'lapan_mv' starts job with 'ADVRES' option to qsub. Job scheduled
and runs cleanly. Before first job terminated, the same user starts
another job, which goes into idle state because of lack of nodes.
When first job termintes, scheduler is disappeared. There is nothing
special is log files (loglevel is 3).
Several times after this, scheduler failed to start again (said
something strange about 'free ()' and halted by segfault).
Can anyone comment this?
If you need more information, don't hesitate to ask -- this problem is
bothering me a lot.
Max Lapan <lapan_mv 'at' inbox 'dot' ru>, +7(0855)296471, ICQ: 233841810
HPC Architect/Administrator, Engineering Analysis Systems Department
JSC Saturn http://www.npo-saturn.ru
More information about the mauiusers