Message boards :
Number crunching :
Ways to bypass the limits.
Message board moderation
Author | Message |
---|---|
Tim Send message Joined: 19 May 99 Posts: 211 Credit: 278,575,259 RAC: 0 |
I was fooling around my tasks and I saw a user with 1400+ tasks in progress. Did he found a way to bypass the limits, or it is something else? I already PM him a week ago, but he didn’t reply. The user is Mr. John Chrzastek and his computer id is 7033569 http://setiathome.berkeley.edu/hosts_user.php?userid=9738701 |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Resend lost tasks is off, he's probably got a flaky connection and getting scheduler timeouts, tasks that never reached his host will time out eventually, Claggy |
Tim Send message Joined: 19 May 99 Posts: 211 Credit: 278,575,259 RAC: 0 |
I will keep an eye on him. Even with lost tasks he must lost and found his files 8 times. And the tasks are still growing. Tim |
S@NL Etienne Dokkum Send message Joined: 11 Jun 99 Posts: 212 Credit: 43,822,095 RAC: 0 |
I don't want to accuse anyone but looking at the tasks he's probably using the rescheduler swapping all MB's to the CPU and getting more and more 100 tasks for his GPU or the other way around... Don't know why because there has hardly been a shortage of MB work. |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
I will keep an eye on him. Even with lost tasks he must lost and found his files 8 times. That's the point, with 'Resend lost tasks' being off, the scheduler can't resend the lost tasks, I recon his client has got a maximum of 124 tasks on hand (probably less), Boinc tells the scheduler it's got, say 40 GPU, and say 20 CPU, But scheduler can't check Boinc database and confirm what tasks client has, so scheduler just sends another say 20 GPU and 10 CPU Wu, still well within limit, But scheduler reply doesn't get to client, So client asks again, saying it's got now 39 GPU, and 20 CPU, scheduler sends some more work, and doesn't get through again, a bit later a scheduler reply does get through, and client gets some fresh work. Claggy |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
I don't want to accuse anyone but looking at the tasks he's probably using the rescheduler swapping all MB's to the CPU and getting more and more 100 tasks for his GPU or the other way around... That's not happening, he's not rescheduling, he has great blocks of lost tasks, all sent at the same time, none of those lost blocks are getting returned as completed by a different application. All tasks for computer 7033569 Claggy |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
I don't want to accuse anyone but looking at the tasks he's probably using the rescheduler swapping all MB's to the CPU and getting more and more 100 tasks for his GPU or the other way around... Following on, these are blocks of work that have been sent, where later blocks have been completed, and these haven't: 13 Mar 2015, 22:17:27 UTC 5 tasks 13 Mar 2015, 22:25:31 UTC 77 tasks 13 Mar 2015, 22:35:55 UTC 137 tasks 13 Mar 2015, 22:47:48 UTC 95 tasks 13 Mar 2015, 23:00:34 UTC 96 tasks 13 Mar 2015, 23:18:47 UTC 23 tasks 14 Mar 2015, 5:42:44 UTC 46 tasks 14 Mar 2015, 7:46:28 UTC 49 tasks 14 Mar 2015, 8:50:42 UTC 24 tasks 15 Mar 2015, 0:22:13 UTC 95 tasks 15 Mar 2015, 0:35:51 UTC 68 tasks 15 Mar 2015, 0:47:52 UTC 69 tasks 15 Mar 2015, 0:55:14 UTC 100 tasks 15 Mar 2015, 1:01:48 UTC 44 tasks 15 Mar 2015, 11:07:06 UTC 10 tasks 15 Mar 2015, 11:54:24 UTC 18 tasks 15 Mar 2015, 12:40:13 UTC 47 tasks 15 Mar 2015, 15:21:23 UTC 48 tasks 15 Mar 2015, 21:53:27 UTC 1 task 16 Mar 2015, 21:48:30 UTC 90 tasks 17 Mar 2015, 0:03:03 UTC 4 tasks 18 Mar 2015, 0:06:29 UTC 5 tasks 18 Mar 2015, 3:48:14 UTC 2 tasks 18 Mar 2015, 5:59:57 UTC 4 tasks 19 Mar 2015, 0:11:29 UTC 10 tasks 19 Mar 2015, 6:50:22 UTC 3 tasks 19 Mar 2015, 23:32:38 UTC 47 tasks 20 Mar 2015, 0:15:14 UTC 13 tasks 20 Mar 2015, 1:54:54 UTC 5 tasks 20 Mar 2015, 11:18:29 UTC 10 tasks 20 Mar 2015, 22:19:06 UTC 6 tasks 20 Mar 2015, 23:39:35 UTC 8 tasks 21 Mar 2015, 0:06:05 UTC 1 task 21 Mar 2015, 2:13:37 UTC 4 tasks 21 Mar 2015, 11:31:13 UTC 71 tasks 21 Mar 2015, 12:48:45 UTC 36 tasks 21 Mar 2015, 14:14:15 UTC 36 tasks 21 Mar 2015, 16:11:04 UTC 19 tasks 21 Mar 2015, 17:30:31 UTC 20 tasks 21 Mar 2015, 18:37:24 UTC 9 tasks 22 Mar 2015, 0:04:09 UTC 29 tasks 22 Mar 2015, 2:07:53 UTC 78 tasks Total 1562 definite lost tasks. 4 tasks definitely received and in progress (2 CPU and 2 GPU) These blocks may or may not have been received (Total 124 including the 4 inprogress Wu's): 22 Mar 2015, 3:39:05 UTC 39 tasks 22 Mar 2015, 5:36:46 UTC 31 tasks Note: there's less than 5 minutes between these tasks been issued, whether these are from one or two scheduler requests is open to question: 22 Mar 2015, 5:38:44 UTC 20 tasks 22 Mar 2015, 6:43:01 UTC 10 tasks 22 Mar 2015, 9:29:28 UTC 20 tasks 1687 minus 1562 makes a possible maximum of 125 tasks in progress, But probably less. Edit: Yes, I know my counting is off by one, I'm not going to go through and manually count the number of Wu's in each block again. Claggy |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
And that's one of the ways that database bloat starts, when 'resend lost tasks' is turned off. There's already the beginning of an uptick visible for 'results in the field' (blue line): That's not enough to account for today's database slowness - that must be something else - but it's ominous. The slower the database is, the more lost results get created: and the excess results in the table indexes slow the database down even more. I don't expect full-blown inflation to set in until next weekend or later, but it could be on the way. |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
I thought too that the resend off would just shoot tasks though the roof with waiting up to 6 weeks for a task that can't be completed to resend, then wait again. I have a feeling it is "Resend OFF and Deleted" since tasks are not accumulating. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
I noticed something similar on a host I was watching. Seems there are quite a few older tasks that are not being worked. While I was watching the older tasks newer tasks were being worked and completed. The best I can figure this host has around 79 'Lost' tasks waiting to time out; http://setiathome.berkeley.edu/results.php?hostid=7159696&offset=360&show_names=0&state=0&appid= I have about a dozen 'Lost' tasks on my hosts, that I know of. It would be nice if 'Resent lost tasks' could be enabled for at least a few hours a week so these tasks could be dealt with rather than waiting to time out. |
bluestar Send message Joined: 5 Sep 12 Posts: 6995 Credit: 2,084,789 RAC: 3 |
Not about limits, but personally I would rather prefer having those tasks carrying out the gaussian search and not those tasks which do not do this thing. Also the .vlar tasks are apparently returning nothing of interest as well. Those tasks having a high angle range could possibly be of more interest, though. |
BilBg Send message Joined: 27 May 07 Posts: 3720 Credit: 9,385,827 RAC: 0 |
Also the .vlar tasks are apparently returning nothing of interest as well. ????? Those tasks having a high angle range could possibly be of more interest, though. ??? P.S. What happened with musicplayer ... ;) Â - ALF - "Find out what you don't do well ..... then don't do it!" :) Â |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.