reducing overhead for quick jobs
Hi All, After changing the sleep in manager.py from 5 seconds to 0.2 seconds, the job start is very snappy. I did not touch the various sleep(10)s that don't seem to be relevant. I'm testing how long it takes to sort a short bed file in galaxy. After the fix, the problem seems to be at the end of the process. In the timeline below the gap between the job ending on the server and the client getting the update is 5 seconds. Is that a built in sleep or something else? Can I turn off metadata for tools that don't generate it? BTW: thousands of seconds are on in the manager and handler logs and should be turned on in the main.log. Rather than 2 round trips to the server. I think we need a special class of tools that run in a single round trip (like the eyeball viz). The max run time for this class of jobs should be no more than a few seconds. Is that is the roadmap somewhere? -Robert Baertsch UC Santa Cruz https://medbook.ucsc.edu RUN1 - 9 seconds (1 sec to start, 0.2 sec to run , 2 sec metadata, 5.5 secs to update client) 12:08:12.? main: log job on main thread 12:08:12.8 manager : galaxy.jobs.manager DEBUG 2013-03-30 12:08:12,867 (2308) Job assigned to handler 'handler2' 12:08:13.26 handler: start handler 12:08:13.35 handler: job dispatched 12:08:13,44 handler: job runner started 12:08:13,72 handler: python tool started 12:08:13,91 handler: python tool finished 12:08:13,98 handler: set_metadata.sh started 12:08:15,70 handler: metadata ended 12:08:15,85 handler: job Ended 12:08:21 main: /api/histories - update on client RUN2 - 9 seconds (1.8 sec to start, 0.2 sec to run, 2 sec metadata, 4.5 sec to update client) 12:23:00.? main: log job on main thread 12:23:00,60 manager : galaxy.jobs.manager DEBUG 2013-03-30 12:08:12,867 (2308) Job assigned to handler 'handler2' 12:23:01,63 handler: start handler 12:23:01,72 handler: job dispatched 12:23:01,81 handler: job runner started 12:23:02,44 handler: python tool started 12:23:02,64 handler: python tool finished 12:23:02,72 handler: set_metadata.sh started 12:23:04,64 handler: metadata ended 12:23:04,71 handler: job Ended 12:23:09 main: /api/histories - update on client
participants (1)
-
Robert Baertsch