climateprediction.net home page
Task loss notification

Task loss notification

Questions and Answers : Getting started : Task loss notification
Message board moderation

To post messages, you must log in.

AuthorMessage
ToKamaK

Send message
Joined: 9 Dec 18
Posts: 2
Credit: 40,027
RAC: 0
Message 59656 - Posted: 20 Feb 2019, 19:49:11 UTC

Good Day,

I wished to notify climateprediction.net one way or another, about the loss of the job #21490082 (work unit #11730305), following a disk crash.

No much problem on my side, thanks to good'ole backup policies; but the I see the deadline set to next year, and wanted to avoid to the submitter of the task to wait a whole year before seing the job being rescheduled.

Is there a proper, practical way to send such notification?
Or is such loss of jobs much of a concern regarding the compute infrastructure?

With my apologies if any delay has been induced,
Kind Regards,
--
ToKamaK
ID: 59656 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 6951
Credit: 20,843,205
RAC: 0
Message 59657 - Posted: 20 Feb 2019, 20:36:31 UTC

There's nothing setup to do this.
The only method to release a model early, is to Abort it. Which isn't possible in your case.

When tasks are taking too long to be returned, the researcher can just close that batch and issue a new one, with a set of tasks to fill in the gaps.

The "deadline" is just a BOINC thing, and it's set way in the future to fool BOINC into not constantly running cpdn tasks instead of switching between the tasks of all other projects as well, which is what happens with a shorter value.
The climate models are actually required ASAP.

These new model, hadam4, have a bug, which is still being worked on.
The only way to keep them running is, under no circumstances let them be interrupted. Which can be done, but it won't suit a lot of people.

For tips on running models under more usual circumstances, see this post on the Care and Feeding of climate models:
Tayloring BOINC settings for cpdn

Matters pertaining to the running of models is usually discussed in Number crunching.
ID: 59657 · Report as offensive     Reply Quote
ToKamaK

Send message
Joined: 9 Dec 18
Posts: 2
Credit: 40,027
RAC: 0
Message 59660 - Posted: 21 Feb 2019, 21:21:31 UTC - in response to Message 59657.  

Good Day,

Thank you for your answer, I get that there isn't much harm done then.

Also, many thanks for the pointer about caring for cpdn tasks, I am running other BOINC projects for quite some time, and the trick about not interrupting jobs is not exactly obvious. I will modify my BOINC configuration accordingly.

Kind Regards
ID: 59660 · Report as offensive     Reply Quote

Questions and Answers : Getting started : Task loss notification

©2019 climateprediction.net