Problem with DCF and BOINC cache

Message boards : BOINC client : Problem with DCF and BOINC cache
Message board moderation

To post messages, you must log in.


Send message
Joined: 16 Jan 08
Posts: 8
Message 22803 - Posted: 31 Jan 2009, 23:20:14 UTC

Recently I have seen some issues with how DCF is managing cache.

This is also occurring on several of my team’s hosts.

When running Milkyway the cache does not stay full - BOINC waits until the very last WU is returned before asking for more work.

I have removed all other projects from the hosts, and it does not affect the problem.

I can go in and manually edit the DCF - work will download, and everything will be fine,, for awhile. Then BOINC will eventfully go back to not asking for work again until the last WU is returned.

After modifying the DCF line to .2xxxxx it works fine, after about 2-3 hours, work is not longer being downloaded and DCF is set back to 0.011 or less.

Not a real problem on single core boxes (if you don’t like a cache!), But the more cores you have, the bigger the problem is. On a 8-Core system, 7 Cores run dry and wait for up to 20 or 30 minutes until the last WU is reported.

What is going on, and how to fix this problem?

Is there a boinccmd line command that can be run as a script to reset the DCF back to normal.

It seems to happen on multiple versions of BOINC, I have tried everything going back to 5.10.30 to the new 6.6.3 all the same problems.

Using optimized app the WU’s take an average of 4-9 minutes to process, and MW only allows a cache of 8 WU per core.

Working example <duration_correction_factor>0.026645</duration_correction_factor>

Failing example <duration_correction_factor>0.013645</duration_correction_factor>

Any ideas?
ID: 22803 · Report as offensive
Profile KSMarksPsych

Send message
Joined: 30 Oct 05
Posts: 1239
United States
Message 22805 - Posted: 1 Feb 2009, 3:41:32 UTC

Milkyway only? Or other projects as well?
Kathryn :o)
ID: 22805 · Report as offensive

Send message
Joined: 8 Jan 06
Posts: 36
United States
Message 22807 - Posted: 1 Feb 2009, 5:43:30 UTC - in response to Message 22803.  
Last modified: 1 Feb 2009, 5:44:31 UTC


Check your logs closer. IIRC, when you hit a max task in progress limit you get a 30 minute comm deferral from MW, which will result in the effect you're seeing.

Given that the opti app is running about twice as fast as the stocker, the TDCF reduction is what I would expect.

ID: 22807 · Report as offensive

Send message
Joined: 6 Dec 06
Posts: 118
United States
Message 22818 - Posted: 1 Feb 2009, 15:23:44 UTC - in response to Message 22805.  

Milkyway only? Or other projects as well?

By my observation, any project will do it when the DCF drops below 0.020000

It happened to me running GeneticLife with cc 5.10.45 some months ago. The client would not request more GL tasks until the GL queue was empty. Just like Kevint, I had to edit client_state and change DCF to something higher than 0.02000 to get the client to request work.
ID: 22818 · Report as offensive

Send message
Joined: 16 Jan 08
Posts: 8
Message 22827 - Posted: 1 Feb 2009, 20:22:31 UTC - in response to Message 22819.  

Milkyway only

there is a discussion on the milkyway forum regarding this. I am not the only one seeing this problem.

BOINC simply does not ask for work until the last WU is processed and returned. There is no delay, and hitting the update button on the BOINC client does not good. BOINC responds with "not asking for more work" statement.

A reset of the project works - for awhile - but soon, within a couple of hours it reverts back to the same issue.

ID: 22827 · Report as offensive

Message boards : BOINC client : Problem with DCF and BOINC cache

Copyright © 2022 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.