climateprediction.net (CPDN) home page
Thread 'Computing error'

Thread 'Computing error'

Questions and Answers : Unix/Linux : Computing error
Message board moderation

To post messages, you must log in.

AuthorMessage
d34th

Send message
Joined: 13 Jul 05
Posts: 1
Credit: 78,204
RAC: 0
Message 17971 - Posted: 10 Dec 2005, 12:23:52 UTC

I\'m running climateprediction under BOINC 5.2.13
I\'m using ubuntu breezy with the i686kernel (intalled from apt)

When the project is running it displays an error (computation error)

And this in the message window

sáb 10 dic 2005 06:33:08 CET|climateprediction.net|Unrecoverable error for result sulphur_fpnp_000733093_0 (process got signal 11)

Greetings from Spain
ID: 17971 · Report as offensive     Reply Quote
Profilegeophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2187
Credit: 64,822,615
RAC: 5,275
Message 17978 - Posted: 10 Dec 2005, 14:26:15 UTC

If you can, I would go to your \"General Preferences\", available from the \"Your account\" link on the left hand navigation bar, and edit them to \"Leave applications in memory when pre-empted\" to Yes.

What happens with sulphur cpdn is that a benchmark occurs, or a switch to another project occurs, and if the cpdn application isn\'t left in memory when that happens, this signal 11 error may show up. This type of problem generally doesn\'t happen in the Windows version. I\"m not sure why it happens in the Linux verison.
ID: 17978 · Report as offensive     Reply Quote
Profileold_user85254

Send message
Joined: 27 Jun 05
Posts: 74
Credit: 199,198
RAC: 0
Message 17995 - Posted: 10 Dec 2005, 17:35:39 UTC - in response to Message 17978.  

If you can, I would go to your \"General Preferences\", available from the \"Your account\" link on the left hand navigation bar, and edit them to \"Leave applications in memory when pre-empted\" to Yes.
.


This is good advice for another reason as well, even without the signal 11 issue

With sulphur you can lose almost an hours work (*) when the app is pre-empted, and as far as I know the benchmarks do not wait for a good time to run so as to avoid this issue. On other projects the apps checkpoint every minute of every few minutes and it is not really an issue; the BOINC defaults make sense on those projects but not for CPDN.

River~~

(*) With slab models it was up to around 20 mins work; sulphur takes around 2x - 3x as long, so I am assuming the checkpoints are also that much further apart
ID: 17995 · Report as offensive     Reply Quote
old_user116053

Send message
Joined: 24 Nov 05
Posts: 1
Credit: 78,689
RAC: 0
Message 18186 - Posted: 14 Dec 2005, 12:46:26 UTC

Hello,

I\'m running climateprediction with BOINC 5.2.13 under SuSE Linux 9.3.

I have the same probs with signal 11 messages, although my preferences are already \"Leave applications in memory when pre-empted - Yes\".

I had 1 model running a few days and got 3 trickles, then it broke. Since then, no model ran longer than until the first switch (I run several other BOINC apps also).

Greetings from Germany

ID: 18186 · Report as offensive     Reply Quote
Desti

Send message
Joined: 6 Aug 04
Posts: 124
Credit: 9,195,838
RAC: 0
Message 18214 - Posted: 14 Dec 2005, 23:16:11 UTC

If you make a daily backup of the complete BOINC dir, you will be able to restart the workunit and you lose only some hours of work.
Linux Users Everywhere @ BOINC
ID: 18214 · Report as offensive     Reply Quote

Questions and Answers : Unix/Linux : Computing error

©2024 cpdn.org