Task computation error gone on reboot

Message boards : BOINC Manager : Task computation error gone on reboot

Author	Message
theducks Send message Joined: Apr 16 08 Posts: 4	Message 16676 - Posted 16 Apr 2008 16:12:31 UTC
	Boinc 5.10.45 win XP pro SP2+ 32bit AMD 2600+ 1.5G RAM Task shows computational error and is now working on next task in list. Re-boot and task list now just shows task completed. (force)Update project project site: my tasks shows Done:error BUG or WAD
	ID: 16676 \|

Sekerob Send message Joined: Aug 25 06 Posts: 1079	Message 16677 - Posted 16 Apr 2008 16:17:11 UTC - in response to Message 16676.
	Boinc 5.10.45 win XP pro SP2+ 32bit AMD 2600+ 1.5G RAM Task shows computational error and is now working on next task in list. Re-boot and task list now just shows task completed. (force)Update project project site: my tasks shows Done:error BUG or WAD Most often the best place to go is the project forum for whom the work failed. So, WAD project BUGS? ____________ Coelum Non Animum Mutant, Qui Trans Mare Currunt
	ID: 16677 \|

Dagorath Send message Joined: Jun 13 07 Posts: 638	Message 16678 - Posted 16 Apr 2008 16:47:20 UTC - in response to Message 16676.
	Boinc 5.10.45 win XP pro SP2+ 32bit AMD 2600+ 1.5G RAM Task shows computational error and is now working on next task in list. Re-boot and task list now just shows task completed. (force)Update project project site: my tasks shows Done:error BUG or WAD No, it's not a bug, just imprecise feedback to user. It would be less confusing if the task list said something like "task completed but with computation error" after a reboot. That would be more congruent with what task list says prior to re-boot. It's not a biggee in my books but others may feel differently.
	ID: 16678 \|

theducks Send message Joined: Apr 16 08 Posts: 4	Message 16679 - Posted 16 Apr 2008 16:48:21 UTC
	I "know" I have a intermittent HW problem (memory bus) that may have caused the error. Since the error was logged to BOINC manager, I assumed it was now the managers job to keep the correct status SHOWING on the User Interface task list through reboots. Rosetta (the site) acknowledged the error that the task originally showed. So how does BOINC manager track result "status" before the result is returned? WAD?
	ID: 16679 \|

theducks Send message Joined: Apr 16 08 Posts: 4	Message 16680 - Posted 16 Apr 2008 16:52:24 UTC - in response to Message 16678.
	Boinc 5.10.45 win XP pro SP2+ 32bit AMD 2600+ 1.5G RAM Task shows computational error and is now working on next task in list. Re-boot and task list now just shows task completed. (force)Update project project site: my tasks shows Done:error BUG or WAD No, it's not a bug, just imprecise feedback to user. It would be less confusing if the task list said something like "task completed but with computation error" after a reboot. That would be more congruent with what task list says prior to re-boot. It's not a biggee in my books but others may feel differently. It would be nice to have correct status in this case rather than let my computer continue to crunch out garbage since my task list has many tasks to do before the automatic return kicks in.
	ID: 16680 \|

Dagorath Send message Joined: Jun 13 07 Posts: 638	Message 16681 - Posted 16 Apr 2008 17:31:31 UTC - in response to Message 16679. Last modified: 16 Apr 2008 17:34:02 UTC
	I "know" I have a intermittent HW problem (memory bus) that may have caused the error. Since the error was logged to BOINC manager, I assumed it was now the managers job to keep the correct status SHOWING on the User Interface task list through reboots. Rosetta (the site) acknowledged the error that the task originally showed. If Rosetta acknowledged it then the manager should acknowledge it all the time, even after a reboot or BOINC restart. So how does BOINC manager track result "status" before the result is returned? WAD? The status of each WU is recorded as an int in client_state.xml until the result is reported successfully. The client writes the status, the manager reads the status and interprets it.
	ID: 16681 \|

Dagorath Send message Joined: Jun 13 07 Posts: 638	Message 16683 - Posted 16 Apr 2008 17:47:26 UTC - in response to Message 16680.
	It would be nice to have correct status in this case rather than let my computer continue to crunch out garbage since my task list has many tasks to do before the automatic return kicks in. I agree. If a task ends with any kind of error then that error should be reported in the task list even after a reboot/restart. Allowing "computation error" to revert to just "complete" is less than user friendly if not misleading.
	ID: 16683 \|

Dagorath Send message Joined: Jun 13 07 Posts: 638	Message 16685 - Posted 16 Apr 2008 18:49:14 UTC - in response to Message 16684.
	Anway, I submitted Trac ticket #621 for this issue.
	ID: 16685 \|

Richard Haselgrove Send message Joined: Oct 5 06 Posts: 443	Message 16686 - Posted 16 Apr 2008 19:51:03 UTC - in response to Message 16685. Last modified: 16 Apr 2008 19:51:20 UTC
	Anway, I submitted Trac ticket #621 for this issue. I think the way the WU outcome is reported on the Tasks page of a project's website also changes the same way: Client error Compute error if the task is reported before a BOINC restart, and Client error Done if it's reported afterwards. That could be fixed at the same time.
	ID: 16686 \|

Dagorath Send message Joined: Jun 13 07 Posts: 638	Message 16688 - Posted 16 Apr 2008 20:47:45 UTC - in response to Message 16686.
	Anway, I submitted Trac ticket #621 for this issue. I think the way the WU outcome is reported on the Tasks page of a project's website also changes the same way: Client error Compute error if the task is reported before a BOINC restart, and Client error Done if it's reported afterwards. I think you may be right but I'm not about to induce a computation error in any of the tasks I have running at the moment to verify (too much CPU time invested). Are you certain? Can anyone else verify? That could be fixed at the same time. That would be nice.
	ID: 16688 \|

Richard Haselgrove Send message Joined: Oct 5 06 Posts: 443	Message 16689 - Posted 16 Apr 2008 21:15:19 UTC - in response to Message 16688.
	I think you may be right but I'm not about to induce a computation error in any of the tasks I have running at the moment to verify (too much CPU time invested). Are you certain? Can anyone else verify? I have a slightly unstable new build which has thrown up a few errors recently. If I see another one, I'll do the restart and see what changes. Of course, by typing that I'm trying to invoke Murphy - when you want an error, everything runs smoothly. But does this count as a double-Murphy?
	ID: 16689 \|

theducks Send message Joined: Apr 16 08 Posts: 4	Message 16690 - Posted 16 Apr 2008 21:21:56 UTC - in response to Message 16685.
	Anway, I submitted Trac ticket #621 for this issue. Thanks. I have been using BOINC for a bunch of years now (Seti,Rosetta and Simap). This is the first time I get to contribute a bug reporting that can do nothing but make BOINC better in the end. Thanks to the BOINC staff (and volunteers) for their effort
	ID: 16690 \|

Richard Haselgrove Send message Joined: Oct 5 06 Posts: 443	Message 16888 - Posted 26 Apr 2008 18:19:33 UTC - in response to Message 16688.
	I think the way the WU outcome is reported on the Tasks page of a project's website also changes the same way: Client error Compute error if the task is reported before a BOINC restart, and Client error Done if it's reported afterwards. I think you may be right but I'm not about to induce a computation error in any of the tasks I have running at the moment to verify (too much CPU time invested). Are you certain? Can anyone else verify? I'm still not certain, but I can supply some corroborative evidence. SETI Beta WU 1105766 is a normal 'surplus to requirements' cancellation. SETI Beta WU 1106057 is the same thing, but with a local BOINC restart between the server issuing the cancellation instruction, and the client reporting that it has received it. We get the "Done" client state, instead of the proper one.
	ID: 16888 \|

Dagorath Send message Joined: Jun 13 07 Posts: 638	Message 16896 - Posted 26 Apr 2008 23:17:36 UTC - in response to Message 16888.
	I think the way the WU outcome is reported on the Tasks page of a project's website also changes the same way: Client error Compute error if the task is reported before a BOINC restart, and Client error Done if it's reported afterwards. I think you may be right but I'm not about to induce a computation error in any of the tasks I have running at the moment to verify (too much CPU time invested). Are you certain? Can anyone else verify? I'm still not certain, but I can supply some corroborative evidence. SETI Beta WU 1105766 is a normal 'surplus to requirements' cancellation. SETI Beta WU 1106057 is the same thing, but with a local BOINC restart between the server issuing the cancellation instruction, and the client reporting that it has received it. We get the "Done" client state, instead of the proper one. Hmmm. Another state reported incorrectly. You may want to add that to the Trac ticket for this issue. The comment Nicolas added to the ticket explains all.
	ID: 16896 \|

Message boards : BOINC Manager : Task computation error gone on reboot

BOINC home page | Log in | Create account

Copyright © 2009 University of California. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.