Advanced search

Message boards : Graphics cards (GPUs) : No work issues - January 2

Author Message
Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1112
Credit: 6,162,416,256
RAC: 0
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5172 - Posted: 2 Jan 2009 | 15:06:04 UTC

It's broken again:

1/2/2009 8:58:42 AM Sending scheduler request: To fetch work. Requesting 950422 seconds of work, reporting 0 completed tasks
1/2/2009 8:58:47 AM Scheduler request completed: got 0 new tasks
1/2/2009 8:58:47 AM Message from server: No work sent
1/2/2009 8:58:47 AM Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.
1/2/2009 8:58:47 AM Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
1/2/2009 8:59:32 AM Sending scheduler request: Requested by user. Requesting 950422 seconds of work, reporting 0 completed tasks
1/2/2009 8:59:37 AM Scheduler request completed: got 0 new tasks
1/2/2009 8:59:37 AM Message from server: No work sent
1/2/2009 8:59:37 AM Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
1/2/2009 8:59:37 AM Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.
1/2/2009 9:00:02 AM Sending scheduler request: Requested by user. Requesting 950422 seconds of work, reporting 0 completed tasks
1/2/2009 9:00:07 AM Scheduler request completed: got 0 new tasks
1/2/2009 9:00:07 AM Message from server: No work sent
1/2/2009 9:00:07 AM Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
1/2/2009 9:00:07 AM Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.

Profile K1atOdessa
Send message
Joined: 25 Feb 08
Posts: 249
Credit: 370,320,941
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5173 - Posted: 2 Jan 2009 | 16:08:59 UTC - in response to Message 5172.

It's broken again:

1/2/2009 8:58:42 AM Sending scheduler request: To fetch work. Requesting 950422 seconds of work, reporting 0 completed tasks
1/2/2009 8:58:47 AM Scheduler request completed: got 0 new tasks
1/2/2009 8:58:47 AM Message from server: No work sent
1/2/2009 8:58:47 AM Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.
1/2/2009 8:58:47 AM Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
1/2/2009 8:59:32 AM Sending scheduler request: Requested by user. Requesting 950422 seconds of work, reporting 0 completed tasks
1/2/2009 8:59:37 AM Scheduler request completed: got 0 new tasks
1/2/2009 8:59:37 AM Message from server: No work sent
1/2/2009 8:59:37 AM Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
1/2/2009 8:59:37 AM Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.
1/2/2009 9:00:02 AM Sending scheduler request: Requested by user. Requesting 950422 seconds of work, reporting 0 completed tasks
1/2/2009 9:00:07 AM Scheduler request completed: got 0 new tasks
1/2/2009 9:00:07 AM Message from server: No work sent
1/2/2009 9:00:07 AM Message from server: Full-atom molecular dynamics for Cell processor is not available for your type of computer.
1/2/2009 9:00:07 AM Message from server: Full-atom molecular dynamics on Cell processor is not available for your type of computer.


I was getting that yesterday, time after time. Just give it 30 minutes and do a manual update. Maybe try every 15 minutes -- after a while I did get new work.

Neil A
Send message
Joined: 9 Oct 08
Posts: 50
Credit: 12,676,739
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwat
Message 5180 - Posted: 3 Jan 2009 | 1:41:54 UTC

Ditto...ran into the same problem and used the same solution.

Neil
____________
Crunching for the benefit of humanity and in memory of my dad and other family members.

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5184 - Posted: 3 Jan 2009 | 6:42:18 UTC - in response to Message 5180.
Last modified: 3 Jan 2009 | 6:42:35 UTC

Ditto...ran into the same problem and used the same solution.

Neil


I cycled a little faster than that and got one after about 15 min or so ... so, not sure if the speed is the key or the persistence ... and I just got another on the second call to GPU Grid so now I have one ... make that two in the queue ...

I think the feeder "clog" issue is back ... where the feeder gets full of one and only one type of task instead of retaining a balance. I suggested several changes years ago ... ah well ... it is sure acting like the queue gets full of tasks for the PS3 and has no room for the tasks for the PCs ... unless they are the same tasks?

Only the hairdresser knows for sure ...

Anyway, I am back running with one in work and two in queue ...

Profile DoctorNow
Avatar
Send message
Joined: 18 Aug 07
Posts: 83
Credit: 122,995,082
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5197 - Posted: 3 Jan 2009 | 15:37:23 UTC
Last modified: 3 Jan 2009 | 15:52:29 UTC

Trying now since over an hour to get a WU with all tricks, no luck... :-\
It gets really annoying with the time.

Edit:
Finally managed to get one, yay.
____________
Member of BOINC@Heidelberg and ATA!

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5202 - Posted: 3 Jan 2009 | 16:29:04 UTC

The computer with the 9800 GT is working on its last, though that is likely to keep it occupied until well into tomorrow ...

The GTC 280 has one more in the queue so it should be running out sometime tonight ... I will have to wait and see if I have to get prayer beads out again ...

In other news, WCG's having server problems for the first time in my memory that has lasted more than a few minutes ... so, I have about 3 days of CEP tasks done and in holding ...

Mind Modeling is also off line ...

The good news, the stress on SIMAP with a high resource share is paying handsome dividends as the growth in the Cobblestones earned is rising nicely (thank you), with a possibility that we may be able to see an increase of as much as 20K in the total.

Other focus projects are likewise seeing more modest gains but even there the increases are heartwarming ...

In sports, the inability to save community preferences continues for more than 20 straight days so you cannot see the glory of my signature ...

Stay tuned ... a special comment may be on the horizon ...

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5225 - Posted: 3 Jan 2009 | 21:31:19 UTC - in response to Message 5184.

I think the feeder "clog" issue is back ... where the feeder gets full of one and only one type of task instead of retaining a balance. I suggested several changes years ago ... ah well ... it is sure acting like the queue gets full of tasks for the PS3 and has no room for the tasks for the PCs ... unless they are the same tasks?


Before the change in the feeder we did not have a single issue which you seem to describe. Sometimes there was no GPU work, but it had just ran dry and the creation of new ones had to be issued manually. There was no strange clogging. And we know the feeder has the bug of not sending work to clients even if there are enough WUs in the queue. So I assume what we're seeing here is just more of the same.

Interesting question though, are the WUs the same? I'd suppose not, but could they be the same? Would one want this? But I digress.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5227 - Posted: 3 Jan 2009 | 21:43:59 UTC - in response to Message 5225.

I think the feeder "clog" issue is back ... where the feeder gets full of one and only one type of task instead of retaining a balance. I suggested several changes years ago ... ah well ... it is sure acting like the queue gets full of tasks for the PS3 and has no room for the tasks for the PCs ... unless they are the same tasks?


Before the change in the feeder we did not have a single issue which you seem to describe. Sometimes there was no GPU work, but it had just ran dry and the creation of new ones had to be issued manually. There was no strange clogging. And we know the feeder has the bug of not sending work to clients even if there are enough WUs in the queue. So I assume what we're seeing here is just more of the same.

Interesting question though, are the WUs the same? I'd suppose not, but could they be the same? Would one want this? But I digress.

MrS


Well, if it would mean I would get work without hassle, then sure I would want them the same. But, I am suspicious that they are not ... some are tagged to go to PS3s and some to the GPU machines. If this is the case then you can get the case where the feeder gets full of PS3 tasks and none for the GPUs ... (or vice versa) ... in this case you have to wait until the PS3 users pull off enough tasks and that some GPU tasks get into the queue.

I forget the project (Rosetta?) but they had this issue where they had tasks of two types and the queue would fill with tasks all of one type and the actual usage was that the other type was pulled faster ... so, you had the problem that if the feeder was pulling in the wrong "mix" of tasks then people were starved.

There was some fiddling with the feeder but no serious work on it at the time ... as far as I know the minor tweak and the fact that the project increased the queue size "solved" the problem. Another case of not actually fixing the problem but just fiddling with the system until it limps along some more ...

What really should be happening in an environment with mixed workloads is that there should be separate queues in the feeder one for each class of work ... then the appropriate list would be queried for the work ... one could argue that this would also allow a shorter total queue length for the sum total of the queues ... but I digress ...

JAMC
Send message
Joined: 16 Nov 08
Posts: 28
Credit: 12,688,454
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwat
Message 5229 - Posted: 3 Jan 2009 | 21:51:12 UTC

I think the number of participants has increased quite dramatically from under 1,000 to over 1,300 in the last two months so unless the number of WU's being issued has increased proportionally this may be a factor as well... ( problems with the SETI cuda program may be sending some this way...)

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 5232 - Posted: 3 Jan 2009 | 22:05:44 UTC - in response to Message 5227.
Last modified: 3 Jan 2009 | 22:06:19 UTC

It would seem natural to use different WUs for PS3 and GPU, especially since until recently we still had the "GPUTEST"-WUs. But using the same WUs for both would certainly benefit project performance due to better load balancing.. if it wouldn't introduce all kinds of new problems.

Regarding the feeder issue.. I still think there are no signs of "the feeder gets full of PS3 tasks and none for the GPUs", but I can't know. I'll forward the question ;)

And, yes, having separate queues is definitely the way to go. That's simple, robust, elegant. Anything else is just tinkering.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 5431 - Posted: 9 Jan 2009 | 23:54:43 UTC - in response to Message 5227.

Dear Paul,
thanks for this suggestion. It was actually the same case for us. BOINC people from Berkeley helped us understanding it the same day that ETA pointed me to your message after my holidays.

thanks.


gdf

Profile Paul D. Buck
Send message
Joined: 9 Jun 08
Posts: 1050
Credit: 37,321,185
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 5447 - Posted: 10 Jan 2009 | 14:05:35 UTC - in response to Message 5431.

Dear Paul,
thanks for this suggestion. It was actually the same case for us. BOINC people from Berkeley helped us understanding it the same day that ETA pointed me to your message after my holidays.

thanks.


gdf


you are most welcome ... just don't tell UCB I was anywhere near your project ... :)

Now, about that 6.56 app ... :)

Post to thread

Message boards : Graphics cards (GPUs) : No work issues - January 2

//