rosetta job never finishes

Questions and Answers : Unix/Linux : rosetta job never finishes

To post messages, you must log in.

AuthorMessage
jgsack

Send message
Joined: 19 Aug 07
Posts: 6
Credit: 176,525
RAC: 0
Message 53849 - Posted: 19 Jun 2008, 18:59:31 UTC

In the last week-or-so (as/of 2008-06-19), I'v been regularly getting rosetta work units getting to 100% but never completing.

In times past this has occurred infrequently and has generally fixed itself on a manager restart, but lately, the restart trick does no good -- it drops back to a about 70-85% and runs up to 100% and hangs there again.

The process doesn't seem to consume cpu cycles, but it prevents other jobs from running.

I was originally running code installed from a shell script, I believe, but upgraded to ubuntu debs since the problem started, so currently I am running boinc-client and boinc-manager 5.10.45 on ubuntu 8.04 (core2duo T7300, 2GB ram). I have seen references to this 100% problem in the context of a "version 5.96".I don't know what v.5.96 is -- could that be related to my problem?

I have aborted my rosetta tasks and suspended rosetta use, so that some other projects can get work done.

What other information can I provide or what else might I try to help resolve this issue.

Regards,
..jim
ID: 53849 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 53878 - Posted: 20 Jun 2008, 18:47:44 UTC

Yes Jim, there have been more Linux problems recently. Especially with t405 tasks. This is discussed on the Number Crunching board, in the various "Problems with..." threads for different versions.
Rosetta Moderator: Mod.Sense
ID: 53878 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
jgsack

Send message
Joined: 19 Aug 07
Posts: 6
Credit: 176,525
RAC: 0
Message 53880 - Posted: 20 Jun 2008, 18:56:37 UTC - in response to Message 53878.  

Yes Jim, there have been more Linux problems recently. Especially with t405 tasks. This is discussed on the Number Crunching board, in the various "Problems with..." threads for different versions.


Let me know if I can provide info, run diagnostics or trial code.

ID: 53880 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
jgsack

Send message
Joined: 19 Aug 07
Posts: 6
Credit: 176,525
RAC: 0
Message 54539 - Posted: 16 Jul 2008, 18:59:04 UTC

After a couple of weeks vacation from Rosetta, I thought I'd give it another try. For the last couple of days (as/of 2008.07.16) the problem of getting stuck without actually finishing has not reoccurred.

Perhaps Ubuntu upgrades have fixed something? Perhaps newer Rosetta task programs have eliminated a bug?

In any case, things seem better now.

I would post my Rosetta version, but I'm not sure how to get it. I do everything from the Boinc Manager, which is version 5.10.45 -- perhaps that's sufficient.

..j
ID: 54539 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 54553 - Posted: 17 Jul 2008, 13:29:33 UTC

Thanks j. Yes, some updates were made that eliminate much of the Linux problems with tasks not running when BOINC thinks they are.

As for test code, this is done via a project called Ralph. Updates to Rosetta are tested there first. You can just attach to it like any other BOINC project, and run it alongside your others. Typically I recommend that you give Ralph a resource share that is about 10% of your Rosetta share.

Be aware that most of the time, Ralph does not have work available. This is generally because the most recent changes have been promoted to Rosetta, and no further changes have been completed to run tests on.
Rosetta Moderator: Mod.Sense
ID: 54553 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Questions and Answers : Unix/Linux : rosetta job never finishes



©2024 University of Washington
https://www.bakerlab.org