| Re: [ptp-dev] LML Driver for SGE/TACC |
| Jeff, If they are omitting parts of qstat output it means they have created a dot file that gives people alternative outputs (likely to prevent people from doing qstat over the whole cluster!) you can still use the original qstat if you know the bin path....for sge you could use something like $SGE_ROOT/bin/$ARCH/qstat (where arch you can get by running [$SGE_ROOT/util/arch] ). I parse from [ qstat -u '*' -xml ] for my users webpage tools (based on PHP) which does include job state. If you're parsing for just the user, qstat -u "username" -xml seems more useful given the thousands of cores. the xml looks like this: <job_list state="running"> <JB_job_number>464793</JB_job_number> <JAT_prio>0.50500</JAT_prio> <JB_name>JobName</JB_name> <JB_owner>Username</JB_owner> <state>r</state> <JAT_start_time>2011-11-29T02:07:01</JAT_start_time> <queue_name>all.q@node113.genus.cluster</queue_name> <slots>8</slots> </job_list> if you want to get a specific queue you could do [ qstat -q all.q -u '*' -xml] To get a list of queues I did it using [ qstat -g c | egrep q ] and this gives me at leas an output showing queue loads (the egrep removes the header... you can do that with awk also... this was easier given my queue.q naming convention) Hope this helps. On Nov 29, 2011, at 10:03 AM, Jeffrey Overbey wrote:
|