| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: Insane short time to complete, tasks get aborted too early |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 18
|
|
| Author |
|
|
adriverhoef
Master Cruncher The Netherlands Joined: Apr 3, 2009 Post Count: 2346 Status: Offline Project Badges: 
|
Woke up this morning to find that the estimated time had dropped to 2 minutes and 59 seconds again. Wondered what would happen with this task. No surprise it got aborted, after 90 minutes, with 106 jobs aboard and sent to the moon after its 101st job.
----------------------------------------
OPNG_0050331_00093_1-- Linux Debian - In Progress 6/12/21 09:51:47 6/15/21 09:51:47 0.00 0.0 / 0.0EDIT: Drastic measures.  From now on I will be aborting received OPNG tasks with too many jobs aboard.[Edit 2 times, last edit by adriverhoef at Jun 12, 2021 8:45:44 PM] |
||
|
|
PMH_UK
Veteran Cruncher UK Joined: Apr 26, 2007 Post Count: 786 Status: Offline Project Badges: 
|
Adri,
----------------------------------------Thanks for the script. The queue on my slow Intel GPUs are a day or more so I need to change every day. Would the below addition to your script correctly change estimate and limit? Having re-checked I probably don't need to change estimate, if anything it is high. # Stop BOINC: sudo systemctl stop boinc-client # Increase the estimated time for OPNG tasks: sudo perl -w -i -p -e ' if (($b,$v,$e) = /(<rsc_fpops_est>)(31449712079576)(\.000000<\/rsc_fpops_est>)/) { $v = 2 * $v; s//$b$v$e/; } if (($b,$v,$e) = /(<rsc_fpops_bound>)(943491362387280)(\.000000<\/rsc_fpops_bound>)/) { $v = 2 * $v; s//$b$v$e/; } ' $CLIENT_STATE # Restart BOINC: sudo systemctl start boinc-client How do you post without formatting getting mangled? Paul.
Paul.
----------------------------------------[Edit 1 times, last edit by PMH_UK at Jun 12, 2021 4:13:07 PM] |
||
|
|
adriverhoef
Master Cruncher The Netherlands Joined: Apr 3, 2009 Post Count: 2346 Status: Offline Project Badges:
|
Hi Paul,
First of all, to answer your question about formatting, you'd have to use the 'Code' button (underneath 'B', 'I', 'U', 'S', 'Size', 'Font' and 'Color', between 'Email', 'Image' and 'List', 'Quote'). Select the -to be formatted- text in your message, click 'Code', presto! Furthermore, you wrote: Adri, If you can put it to use, than it sounds great to me.Thanks for the script. The queue on my slow Intel GPUs are a day or more so I need to change every day. My queues are about one day, too; the only problematic system, concerning estimated time, is a laptop.Would the below addition to your script correctly change estimate and limit? I've tested it without the -i option and it works as expected. So, you did a good job there!  With formatting codes it would look like: # Stop BOINC: |
||
|
|
Crystal Pellet
Veteran Cruncher Joined: May 21, 2008 Post Count: 1403 Status: Offline Project Badges: 
|
Is it just coïncidence?
All 4 tasks of batch 53k I got, are all over 100 inside jobs: 117, 132, 138 and 118 jobs. |
||
|
|
adriverhoef
Master Cruncher The Netherlands Joined: Apr 3, 2009 Post Count: 2346 Status: Offline Project Badges:
|
Sigh. It's happening again: too many jobs in an OPNG-task. Two OPNG-tasks have already been aborted over the past day by the BOINC-client because of the 'exceeded elapsed time limit 7028.48 (1257988.48G/178.98G)' error message after only 1 hour and 55 minutes of execution time.
----------------------------------------The first one that was Lo and behold, a few hours later an OPNG-task with 74 tasks inside already got I guess I will be aborting all OPNG-tasks with too many inside jobs then again, like I did a half year ago, because every new task that arrives is able to reset the estimated time to three minutes or even less. [Edit 2 times, last edit by adriverhoef at Jan 13, 2022 1:14:44 AM] |
||
|
|
alanb1951
Veteran Cruncher Joined: Jan 20, 2006 Post Count: 1317 Status: Offline Project Badges: 
|
Yup, Adri -- the average number of jobs in tasks I've received since the start of 11th January has been over 80 - fortunately I have GPUs that can get through the work in under half an hour :-)
It would probably be an idea for the plan class for Intel GPUs to be designed to cut out the slow and ancient if that's possible. Alternatively, perhaps a cap on the number of jobs in a task (as they did for OPN1 when it had lots of little jobs per task and caused havoc on some systems...) I've just tried to use "Contact us" to ask if they can cap the jobs per task, but I don't know whether it got through -- there's no acknowledgment given on hitting Submit, it just returns to the "Contact us" page with all the fields blank! We shall see... Cheers - Al. |
||
|
|
Grumpy Swede
Master Cruncher Svíþjóð Joined: Apr 10, 2020 Post Count: 2494 Status: Offline Project Badges: 
|
I can crunch these huge WU's without a problem on my old and slow GTX660M.
----------------------------------------They do take 5-6 hours, but they finish just fine. This WU contains 95 jobs, but the old GPU crunched it without any problems: https://www.worldcommunitygrid.org/contribution/workunit/112536555 Added: I'll let my Intel HD4600 crunch a few of them too, just to check it out. But previously when we had high jobs WU's, the HD4600 didn't have any problems to finish them. [Edit 1 times, last edit by Grumpy Swede at Jan 13, 2022 1:37:09 AM] |
||
|
|
Keith Myers
Senior Cruncher USA Joined: Apr 6, 2021 Post Count: 193 Status: Offline Project Badges: 
|
All of my errored tasks get 70-80% finished of the jobs contained in the work unit.
----------------------------------------They fail around job #70 or so usually.  A proud member of the OFA (Old Farts Association) |
||
|
|
|