This request was originally posted on the GPUGrid forums. Over there they believe it could be a problem with BAM! causing the detachments.

Can anyone explain why I am getting over half of my WUs being returned with a status of "Client Detached"?

I am not manually detaching the WUs. Most of them seem to be detached after about 30 minutes and then I get another two WU and sometimes I get to processes them and other times they also get detached.

The system information is as follows:
CPU type GenuineIntel Intel(R) Core(TM) i5 CPU 650 @ 3.20GHz [Family 6 Model 37 Stepping 2]
Number of processors 4
Coprocessors NVIDIA GeForce GTX 460 (993MB ) driver: 25896
Operating System Microsoft Windows 7 x64 Edition, (06.01.7600.00)
BOINC client version 6.10.58
Memory 4087.49 MB
Cache 256 KB
Swap space 8173.13 MB
Total disk space 698.54 GB
Free Disk Space 665.58 GB

This is only happening to GPUGrid and not the other projects I have running on the same host. It started November 27 and has been getting worse the last week or so. Of course it has been working properly the last couple of days now that I am trying to solve this problem

My BAM! configuration appears to be the same for all the projects I have running.

Any assistance would be greatly appreciated.

Thanks

Could you provide a link to the WUs in question?

http://www.gpugrid.net/results.php?hostid=81432&offset=0&show_names=0&state=5





Tough to diagnose with so little physical evidence to go on. But either your client is actually detaching from the project or the project thinks your client is detaching. The best clues will be in your BOINC Manager message log (assuming you still have them after the event). Search messages around the actual time the WU is reported by the project as being returned.

Thanks Magyarficko for your help.

Unfortunately I have lost the messages during the times that WUs were reported as detached. I do remember looking at the time but did not keep any of the messages. I remember seeing messages similar to this but I noticed that there seemed to be no activity at the specific times of the claimed detachments. I will keep my eye out for more problems like this and save all the messages I can the next time I notice a problem.

Hopefully it was just an intermittent issue and won't come back.

11/12/2010 11:54:26 GPUGRID Sending scheduler request: Requested by project.
11/12/2010 11:54:26 GPUGRID Requesting new tasks for CPU and GPU
11/12/2010 11:54:32 GPUGRID Scheduler request completed: got 0 new tasks
11/12/2010 11:54:32 GPUGRID Message from server: No work sent
11/12/2010 11:54:32 GPUGRID Message from server: Fermi-class GPU not supported by cuda2.2
11/12/2010 11:54:32 GPUGRID Message from server: (reached limit of 2 GPU tasks in progress)
11/12/2010 14:49:12 Contacting account manager at http://bam.boincstats.com/
11/12/2010 14:49:15 Account manager: BAM! User-ID: 29497
11/12/2010 14:49:15 Account manager: BAM! Host-ID: 240462
11/12/2010 14:49:15 Account manager: Number of BAM! connections for this host: 224
11/12/2010 14:49:15 Account manager contact succeeded
11/12/2010 15:14:12 GPUGRID Sending scheduler request: To fetch work.
11/12/2010 15:14:12 GPUGRID Requesting new tasks for CPU
11/12/2010 15:14:16 GPUGRID Scheduler request completed: got 0 new tasks
11/12/2010 15:14:16 GPUGRID Message from server: No work sent
11/12/2010 15:14:16 GPUGRID Message from server: Fermi-class GPU not supported by cuda2.2
11/12/2010 15:14:16 GPUGRID Message from server: (reached limit of 2 GPU tasks in progress)
11/12/2010 16:35:22 Project communication failed: attempting access to reference site
11/12/2010 16:35:24 Internet access OK - project servers may be temporarily down.
11/12/2010 20:14:19 GPUGRID Sending scheduler request: Requested by project.
11/12/2010 20:14:19 GPUGRID Reporting 1 completed tasks, requesting new tasks for CPU and GPU
11/12/2010 20:14:27 GPUGRID Scheduler request completed: got 1 new tasks
11/12/2010 20:14:29 GPUGRID Started download of 426-KASHIF_HIVPR_n1_unbound_so_ba1-84-LICENSE
11/12/2010 20:14:29 GPUGRID Started download of 426-KASHIF_HIVPR_n1_unbound_so_ba1-84-COPYRIGHT
11/12/2010 20:14:30 GPUGRID Finished download of 426-KASHIF_HIVPR_n1_unbound_so_ba1-84-COPYRIGHT
11/12/2010 20:14:30 GPUGRID Started download of 426-KASHIF_HIVPR_n1_unbound_so_ba1-84-426-KASHIF_HIVPR_n1_unbound_so_ba1-83-100-RND7516_1


Again, Thanks

jnargus

Well, good luck to you! If you post messages from your log again, let us know your Offset from UTC so we can match messages from your log with tasks on your task list at the project.

I do see one thing I find kind of strange in what you posted (unless there are messages left out). At 15:14 there was contact with the scheduler, no completed tasks being reported. Then, 5 hours later at 20:14 there was another contact with the project scheduler, this time reporting one completed task. Yet between those two times I see no message that any task finished processing ????

Do you know that all messages from the Messages tab are kept in stdoutdae.txt up to a predefined file size? After that, they get swapped into stdoutdae.old. The suffices are for windows, but the base name should be the same for all platforms. The files can be found in the BOINC data directory.

Gruß,
Gundolf