First big post-outage sign of noon stats life:

10/15/10 0:006:14:34:29 15,835 29

and counting. 12 left in PV, double the pre-outage number, but falling.

Seen Ingleside multiple times refer to client auto-offs up to 4 hours ** and knreed posting yesterday to have set the server to boot the clients 1 hour.

Would hope it does not go to 24 hours unless a client has used it's quota due error and asking again same day being pointless. Would be a potential for many clients idling along when work is available.

edit: spelling

edit2: ** The reference discussion http://www.worldcommunitygrid.org/forums/wcg/...ead,30042_offset,0#298081

This is a BOINC client-feature, not server-side, so WCG and GPUGrid works the same way...

If a BOINC projects #uploads (or really, if #tasks with upload-files) is more than 2* #cpus, or 2* #gpu's (whatever is larger), the BOINC client refuses to make any work-request to the affected project.

It doesn't matter if the uploads is currently transferring, or is currently backed-off, it only goes on how many total there is.

Many projects has a single upload per task, in this case it's just to count #files in transfer-tab. So, if example you've got a quad-core, if you've got example 2 uploads in progress and 6 waiting, you've got a total of 8, and can ask for work. If you've got 2 + 7 on the other hand it's 9 total, > 8, and can't ask for work.

For WCG, you'll need to count tasks instead, if example a quad-core, you can have max 8 tasks marked as "Uploading". If these tasks has 8 uploads or 80 doesn't matter, you'll still allowed to ask for more work. With 9 tasks on the other hand, work-request is blocked.


This feature was added to guard against computers that manages crunching tasks faster than they're managing to upload them. If these computers is allowed to continue downloading more work, they'll "soon" return all work after the deadline, and for most projects this would mean the result is useless, and won't give any credit.

As an added bonus it also eases-off the load on project-servers, in cases like now there is a server-problem, but not everyone understands this is an advantage...

I'm out of work. I keep getting "Backing off X hours".

Yesterday I got "Project down for maintanance".

Is this a client-side problem?

PS. the problem is happening NOW, as we speak.

PSS. My quad-core laptop keeps running out of work. It completes most tasks in 2-3 hours then I get the "back-off" message.

All my "slower" computers are crunching away.

Still having the same problem as 3 days ago. I'm using BOINC manager to do WGC tasks...

I tried to "reset" project. It downloaded some new tasks and completed them, but after 3-4 hours I got the same problem. It hasnt even reported the task it completed. They just sit there saying "Ready to report". One of the tasks says "downloading" and it's been like that for 24 hours now.

Here's some of the abnormal messages from my log:
"Temporary failed download of XXX"
"Backing of X hours of task XXX"
"[error] reported by file upload server: Can't open file."

These messages are all over my log.