I have set my preferences to keep enough work on disk to last between 12 and 13 days in general preferences and 10 to 12 days for the home venue. All my computers are set to "home." I am pretty sure that Boinc is using the home preferences, but not positive because of the following messages:
2004-07-27 06:34:59 [---] General prefs: from SETI@home (last modified 2004-07-24 06:29:39)
2004-07-27 06:34:59 [---] General prefs: using your defaults
In any case, here is a table of the names on my computers, the average time per WU, the WUs on hand, and the predicted days supply on hand computed as average time per WU times # of WUs.

7/28/2004 Isaac Luke Moses Laptop Paul Nat Abe
Average 3.14 1.69 3.52 3.64 3.20 3.43 3.57
StdDev 0.19 0.33 0.28 0.38 0.31 0.37 0.60
WUs 55 122 47 51 91 54 72
Days 6.67 15.27 6.59 7.39 11.25 7.4 10.27

Note that Luke has 122 WUs in inventory, enough for 15.27 days, but Isaac has only 55 WUs for 6.67 days supply. The real problem is that Luke, which already has too many WUs, is still trying to download more, while Isaac, which has too few, is not and cannot be made to download a WU (by clicking on update).

First I am going to assume that you have joined another project, probably Beta, sometime because the symptom that Luke is showing fits that.

In this case there is a bug where every time that the Beta project is contacted for an update, it gives a master file fetch error, and downloads another WU from some other project.

The other part of the equation is how much work BOINC thinks that you have. If the estimates that BOINC has for the WUs on your computers total above 10 days, then BOINC will not download more until its estimate is less than that. Some of the WUs a couple of weeks ago were let out with extremely long crunch estimates. Many people dumped these WUs, and they have been sent out for re-processing. If you have a couple of these in your queue, it will show as a reduced WU count for that machine.

If you are going to join Predictor, you should reduce your cache size, as some of the WUs from predictor have deadlines of 1 day.

> First I am going to assume that you have joined another project, probably
> Beta, sometime because the symptom that Luke is showing fits that.
>
> In this case there is a bug where every time that the Beta project is
> contacted for an update, it gives a master file fetch error, and downloads
> another WU from some other project.
>
> The other part of the equation is how much work BOINC thinks that you have.
> If the estimates that BOINC has for the WUs on your computers total above 10
> days, then BOINC will not download more until its estimate is less than that.
> Some of the WUs a couple of weeks ago were let out with extremely long crunch
> estimates. Many people dumped these WUs, and they have been sent out for
> re-processing. If you have a couple of these in your queue, it will show as a
> reduced WU count for that machine.
>
> If you are going to join Predictor, you should reduce your cache size, as some
> of the WUs from predictor have deadlines of 1 day.
>
>
BOINC also adjusts the queue based on active fraction. So if a computer is only running BOINC half the time it will only appear to half fill the queue. If you ran beta for awhile then quit it may make this fraction abnormally low and it will straighten itself out eventually.

John Keck
BOINCing since 2002/12/08