It sounds like the scheduler is "stuck" did you stop and restart? When you restart make sure the -R switch is on (/etc/init.d/fossology). What I've seen happen is the scheduler get's confused and doesn't see any jobs to run. This is one of a handful of reasons the scheduler is on my "hit list" but I won't have time to work on that until 1.2 is done.
Yes the current, in progress, 1.2 is in the svn trunk. But before you do this make sure Mark chimes in since he is testing the trunk. He doesn't seem to be on IRC yet. Bob On Feb 17, 2010, at 1:08 AM, Mike Kinghan wrote: Hi Bob, Mary This is a 2-core machine and I'm using the Scheduler.conf as installed. The %Host lines says 'localhost 1 1'. I've just been watching the bsam-engine process (following Mary's tip) instead of the scheduler. In 2 minutes I didn't see it do anything but sleep. Then I looked at the CPU usage and both cores are dawdling in the range 10-30% usage. If the Elapsed Scheduled/Running numbers are anywhere near accurate, then the license agent just hardly ever runs. I'll move up to v1.2 and maybe all this will become academic. Is v1.2 the current SVN trunk? BR, Mike On 16 February 2010 22:30, Gobeille, Robert <bob.gobei...@hp.com<mailto:bob.gobei...@hp.com>> wrote: Hi Mike, Around 90% of that sleep time has been removed in v 1.2. v1.2 also has a new license detector that runs around 20x faster. It runs so fast that the scheduler cpu time (not sleep time) becomes the bottleneck. I'm hoping to rewrite the scheduler for v1.3. How many license (bsam) threads are you running (in Scheduler.conf)? We typically use the number of cores - 1 (we don't have any machines with a single core). Bob Gobeille On Feb 16, 2010, at 12:24 PM, Laser, Mary wrote: Hi Mike, The fossology scheduler is most likely NOT the limiting factor. During analysis, 99-100% of the CPU is consumed by the process(es) labeled bsam. Depending on how you configured Scheduler.conf file, there could be multiple bsam processes. However, we strongly recommend you do not exceed the number of CPUs available. You can use the mkschedconf utility to create a Scheduler.conf file as described here http://fossology.org/1.0.0:multi-system_setup#part_2the_scheduler. If you are running on a single system, you can simply run: /usr/lib/fossology/mkschedconf -L Mary ________________________________ From: fossology-boun...@fossology.org<mailto:fossology-boun...@fossology.org><mailto:fossology-boun...@fossology.org<mailto:fossology-boun...@fossology.org>> [mailto:fossology-boun...@fossology.org<mailto:fossology-boun...@fossology.org>] On Behalf Of Mike Kinghan Sent: Tuesday, February 16, 2010 3:24 AM To: fossology@fossology.org<mailto:fossology@fossology.org><mailto:fossology@fossology.org<mailto:fossology@fossology.org>> Subject: [FOSSology] How can I get the scheduler to use more CPU time? Hi fossologists, I'm running an experimental fossology 1.1 job on my Ubuntu 9.10 work machine to analyse the Symbian kernel, ~7500 files. This job has been ongoing, whenever the machine is powered on, for > 9 days now and is about 25% through the license analysis stage. This is way too long to be useful - but its only taking so long because the scheduler gives it so little CPU time. The Elapsed Scheduled figure for the license stage is now merely 2hr 27min and Elapsed Running is 2hr 18min. I've spent as long as I can stand watching the scheduler in my processor monitor and have only once for a fleeting fraction of a second caught it not sleeping. I can't find out how to configure the scheduler to be greedier. How do I do this? Cheers, Mike -- Mike Kinghan, Test Lead, Symbian +44(0)776 5222 793 -- Mike Kinghan, Test Lead, Symbian +44(0)776 5222 793 _______________________________________________ fossology mailing list fossology@fossology.org http://fossology.org/mailman/listinfo/fossology
