General prefs are set at "keep between .25 - 1 days work", cache has about 20 predictor@home WUs at over 2 hours each (well over 1 day) and BOINC made a request to Seti@home for "1 second of work". This request was made exactly 1 minute after a predictor RPC was initiated and failed. Similar problem with downloading too many predictor WUs when BOINC Beta test can not be reached for RPC. See below for log. Intel Celeron 2.0 gHz WinXP SP1. BOINC 3.20.

Predictor@home - 2004-07-23 22:35:08 - Sending request to scheduler: http://predictor.scripps.edu/predictor_cgi/cgi
Predictor@home - 2004-07-23 22:35:11 - Scheduler RPC to http://predictor.scripps.edu/predictor_cgi/cgi succeeded
Predictor@home - 2004-07-23 22:35:11 - Message from server: Project is temporarily shut down for maintenance
Predictor@home - 2004-07-23 22:35:11 - Project is down
Predictor@home - 2004-07-23 22:35:11 - Deferring communication with project for 1 minutes and 0 seconds
SETI@home - 2004-07-23 22:36:11 - Requesting 1 seconds of work
SETI@home - 2004-07-23 22:36:12 - Sending request to scheduler: http://setiboincdata.ssl.berkeley.edu/sah_cgi/cgi
SETI@home - 2004-07-23 22:36:16 - Scheduler RPC to http://setiboincdata.ssl.berkeley.edu/sah_cgi/cgi succeeded
SETI@home - 2004-07-23 22:36:16 - Message from server: No work available
--- - 2004-07-23 22:36:16 - General prefs: from Predictor@home (last modified 2004-07-04 21:38:11)
--- - 2004-07-23 22:36:16 - General prefs: using your defaults
SETI@home - 2004-07-23 22:36:16 - No work from project
SETI@home - 2004-07-23 22:36:19 - Deferring communication with project for 57 seconds
Predictor@home - 2004-07-23 22:36:21 - Sending request to scheduler: http://predictor.scripps.edu/predictor_cgi/cgi
Predictor@home - 2004-07-23 22:36:24 - Scheduler RPC to http://predictor.scripps.edu/predictor_cgi/cgi succeeded
Predictor@home - 2004-07-23 22:36:24 - Message from server: Project is temporarily shut down for maintenance
Predictor@home - 2004-07-23 22:36:24 - Project is down
Predictor@home - 2004-07-23 22:36:24 - Deferring communication with project for 1 hours, 0 minutes, and 0 seconds

Yes, this is a known bug. I was hoping it would be fixed in 3.20.

Hopefully 4.x kills it...

<a> [/url]

> Yes, this is a known bug. I was hoping it would be fixed in 3.20.
>
> Hopefully 4.x kills it...
>
> <a> [/url]
>
I think it is still there in CCv4.01 with only one project this bug doesn't show up as well though.

Another thing to consider is that the client attempts to fill until the queue is past the high water mark. So there should be a fraction of a workunit that is beyond the max.

John Keck
BOINCing since 2002/12/08