Computation Errors

Questions and Answers : Web site : Computation Errors

To post messages, you must log in.

AuthorMessage
Destroyer_Kahn

Send message
Joined: 13 Oct 05
Posts: 9
Credit: 63,540
RAC: 0
Message 9585 - Posted: 22 Jan 2006, 14:33:39 UTC

Everything seems to be running fine, but occasionally I see "Computation Error" in the "Status" column on the "Work" tab. Happened yesterday again. It's gone now, I assume it cleared itslef. Is there something I should do to keep this from happening?

Windows XP, Pentium 4 class, BOINC 5.2.13
ID: 9585 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Nightbird

Send message
Joined: 17 Sep 05
Posts: 70
Credit: 32,418
RAC: 0
Message 9586 - Posted: 22 Jan 2006, 15:07:00 UTC

So the wu is invalid ?


ID: 9586 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Destroyer_Kahn

Send message
Joined: 13 Oct 05
Posts: 9
Credit: 63,540
RAC: 0
Message 9607 - Posted: 22 Jan 2006, 21:17:56 UTC - in response to Message 9586.  

So the wu is invalid ?



No idea. All I know is sometimes I see that there is a Computational Error, and then later it is gone.
ID: 9607 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Snake Doctor
Avatar

Send message
Joined: 17 Sep 05
Posts: 182
Credit: 6,401,938
RAC: 0
Message 9609 - Posted: 23 Jan 2006, 0:11:08 UTC - in response to Message 9607.  

So the wu is invalid ?



No idea. All I know is sometimes I see that there is a Computational Error, and then later it is gone.


Just looking at your results files, it looks as though the computation errors you are getting will not get credit. When you say that they are there and then they are gone, i assume you mean from the Work Unit queue on your system in the BOINC manager display. If that is the case, what you are seeing is normal behavior. When the WU errors out it will be listed in the work Queue as a client error or a computation error until the system "reports" its results. When it reports, the listing for the WU on your system will disappear. After that the only place you can see the error is in your stats here at the project web site.

They will be listed there for a while until the project cleans its data base, which happens every few days to a few weeks, then they will be gone from there as well.

The errors you are getting are a bit difficult to diagnose, but based on the error information it seems as though they may be related to a known bug. The fix for the bug is to set your preferences to "keep application in memory durring swaps" to YES. In effect if your are running more than one project this would keep the rosetta application in memory when the other project is active and vise-versa.

If you are only running one project, then you should avoid shutting BOINC down, suspending active work units, suspending project activity for R@H, or rebooting your system. All of these activities interupt the work unit in progress and can cause errors when the WU restarts.

As I said the project team is aware of the bug and they are working on a fix.

Regards
Phil


We Must look for intelligent life on other planets as,
it is becoming increasingly apparent we will not find any on our own.
ID: 9609 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
JKE

Send message
Joined: 3 Mar 06
Posts: 2
Credit: 104,089
RAC: 0
Message 11933 - Posted: 12 Mar 2006, 11:16:11 UTC - in response to Message 9609.  

So the wu is invalid ?



No idea. All I know is sometimes I see that there is a Computational Error, and then later it is gone.


Just looking at your results files, it looks as though the computation errors you are getting will not get credit. When you say that they are there and then they are gone, i assume you mean from the Work Unit queue on your system in the BOINC manager display. If that is the case, what you are seeing is normal behavior. When the WU errors out it will be listed in the work Queue as a client error or a computation error until the system "reports" its results. When it reports, the listing for the WU on your system will disappear. After that the only place you can see the error is in your stats here at the project web site.

They will be listed there for a while until the project cleans its data base, which happens every few days to a few weeks, then they will be gone from there as well.

The errors you are getting are a bit difficult to diagnose, but based on the error information it seems as though they may be related to a known bug. The fix for the bug is to set your preferences to "keep application in memory durring swaps" to YES. In effect if your are running more than one project this would keep the rosetta application in memory when the other project is active and vise-versa.

If you are only running one project, then you should avoid shutting BOINC down, suspending active work units, suspending project activity for R@H, or rebooting your system. All of these activities interupt the work unit in progress and can cause errors when the WU restarts.

As I said the project team is aware of the bug and they are working on a fix.

Regards
Phil



When setting "keep application in memory durring swaps" to yes, what kind of memory is consumed? RAM, harddrive or other?

Thanks,

Jonas
ID: 11933 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Moderator9
Volunteer moderator

Send message
Joined: 22 Jan 06
Posts: 1014
Credit: 0
RAC: 0
Message 11950 - Posted: 12 Mar 2006, 18:49:31 UTC - in response to Message 11933.  

So the wu is invalid ?



No idea. All I know is sometimes I see that there is a Computational Error, and then later it is gone.


Just looking at your results files, it looks as though the computation errors you are getting will not get credit. When you say that they are there and then they are gone, i assume you mean from the Work Unit queue on your system in the BOINC manager display. If that is the case, what you are seeing is normal behavior. When the WU errors out it will be listed in the work Queue as a client error or a computation error until the system "reports" its results. When it reports, the listing for the WU on your system will disappear. After that the only place you can see the error is in your stats here at the project web site.

They will be listed there for a while until the project cleans its data base, which happens every few days to a few weeks, then they will be gone from there as well.

The errors you are getting are a bit difficult to diagnose, but based on the error information it seems as though they may be related to a known bug. The fix for the bug is to set your preferences to "keep application in memory during swaps" to YES. In effect if your are running more than one project this would keep the rosetta application in memory when the other project is active and vise-versa.

If you are only running one project, then you should avoid shutting BOINC down, suspending active work units, suspending project activity for R@H, or rebooting your system. All of these activities interrupt the work unit in progress and can cause errors when the WU restarts.

As I said the project team is aware of the bug and they are working on a fix.

Regards
Phil



When setting "keep application in memory during swaps" to yes, what kind of memory is consumed? RAM, harddrive or other?

Thanks,

Jonas


The application will be moved to your virtual memory during an application swap. It will then be put back in RAM when Roestta starts up again. In addition to the above, you should consider setting the time between application swaps to something like 120 min to provide enough time for Rosetta to "Checkpoint" between swaps. for more detail on this and other related topics please take some time to browse the FAQs thread linked from my signature.

Moderator9
ROSETTA@home FAQ
Moderator Contact
ID: 11950 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
JKE

Send message
Joined: 3 Mar 06
Posts: 2
Credit: 104,089
RAC: 0
Message 11982 - Posted: 13 Mar 2006, 16:56:59 UTC - in response to Message 11950.  

Thanks
ID: 11982 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Questions and Answers : Web site : Computation Errors



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