Work restarting from scratch on reboot.

Questions and Answers : Windows : Work restarting from scratch on reboot.

To post messages, you must log in.

AuthorMessage
Brian S. Wilson

Send message
Joined: 2 Dec 12
Posts: 4
Credit: 97,591
RAC: 0
Message 74959 - Posted: 24 Jan 2013, 3:31:40 UTC
Last modified: 24 Jan 2013, 3:36:04 UTC

My system has to restart as certain times and I suspend the Rosetta@home project then do the restart. All work (many hours of computations)that was done is gone when the system restarts. I thought the project was supposed to write their data every so often so a restart wouldn't lose everything. Any ideas on what the issue might be? I'll check back to see responses as my email isn't working well at the moment.

I've got a Windows XP Pro 32 Bit (SP3) system with 4GB memory, plenty of free disk space, and /temp and my browser cache on a ram drive.
ID: 74959 · 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 74969 - Posted: 24 Jan 2013, 20:49:52 UTC

There is a setting in the compute preferences for "task checkpoint to disk at most every ... seconds". This value is intended to avoid tasks utilizing too much disk IO. You cannot force an application to take a checkpoint, but if you entered a very large value in that entry, you are perhaps preventing it from checkpointing when it would be able to.

Typically tasks checkpoint several times an hour, but it depends on the type of task. Some can take over an hour to reach a point that a checkpoint can be recorded. You can look at the properties of each task and see the current CPU time and the CPU time at which the last checkpoint was taken.

Other than leaving the machine on for longer periods of time, there really isn't much else you can do to effect the behavior. Every time you exit BOINC or power off, you will lose some work, it's just a question of how much.
Rosetta Moderator: Mod.Sense
ID: 74969 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Brian S. Wilson

Send message
Joined: 2 Dec 12
Posts: 4
Credit: 97,591
RAC: 0
Message 74974 - Posted: 25 Jan 2013, 0:43:32 UTC - in response to Message 74969.  

There is a setting in the compute preferences for "task checkpoint to disk at most every ... seconds".


Yes, I apologize for not including this value in my earlier issue statement. It is set to 60 seconds on my system.
ID: 74974 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Brian S. Wilson

Send message
Joined: 2 Dec 12
Posts: 4
Credit: 97,591
RAC: 0
Message 74983 - Posted: 26 Jan 2013, 14:26:07 UTC - in response to Message 74969.  

Typically tasks checkpoint several times an hour, but it depends on the type of task. Some can take over an hour to reach a point that a checkpoint can be recorded. You can look at the properties of each task and see the current CPU time and the CPU time at which the last checkpoint was taken.


I was wondering where the checkpoints are saved. I've always been assuming it was on my C: drive (as I've got 10GB allocated for BOINC); but I don't see where this is specified in the preferences and I don't know where the various configuration files are kept, what they are called, or if they are human readable.
ID: 74983 · 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 74984 - Posted: 26 Jan 2013, 15:03:44 UTC
Last modified: 26 Jan 2013, 15:04:00 UTC

When you installed BOINC it prompted you for a path name under which all of the applications and tasks are stored. You can see the path that was select in the first set of messages in the log file as BOINC starts. Under that path, there will be a slots directory. A "slot" is a task that is in progress. Each task has it's own slot until it has ended. Since checkpoints are associated with specific tasks, they are stored in the slot directory for the task they pertain to.
Rosetta Moderator: Mod.Sense
ID: 74984 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
vakobo

Send message
Joined: 3 Aug 08
Posts: 18
Credit: 13,910,889
RAC: 857
Message 75563 - Posted: 4 May 2013, 18:06:26 UTC

I have default setting for task checkpoints (60 s), but some tasks do not have checkpoint when i see information about those tasks. If i shutdown my computer and turn it on next time those tasks will start from 0%. This is very sad. Especially when task was more then 90% ready.
ID: 75563 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Questions and Answers : Windows : Work restarting from scratch on reboot.



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