That one is short. However, your image is probably appropriate for this one: http://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,34590

twilyth, not yet time for any one to dock in a corner like that and take cover. We have about 3-days and counting down (as I type) before one may do so, else...

Machines may be loaded and cached with enough WUs for 48-days hrs (which is twice the announced expected outage-duration of 24hrs). It's the supply of HCC-v7.08-GPU-WUs that needs to be re-stocked more than other project-WUs. I'm confident that the WCG-servers are prep-ed up for that interim task.
;
; andzgridPost#808
;

Think we explained what could happen when you set the cache higher for HCC [or any other science] than the deadline [7 days]. Watch [total] panic evolve in the task scheduling, which on the later 7.0.4x clients is dealt with much more graceful than with 6.10, but there will be panic. Testing, I've had 8 on the octo running in high priority, *until* after fully caching up, you set the MinB back to what it's normal. BTT the tasks furthest back in the queue get their turn, they will process high priority anyhow [they are then 6-7 days old already]. Now got a bunch due in 4 days, much more than the client can do in 24 hours and the MinB is at 2 days... all are processing High Priority. Best *not* watch BOINC for longer as it's unnerving. If it is, set MinB back to zero days, and the HP state will go away, for longer.

MinB = Minimum work buffer which in the old clients is called "connect about every ... days"

Be warned, that if the server senses you wont be returning work in time per it's logic [though the brighter Homo Sapiens Sapiens at times does], it wont send work. You risk to get the "device on 99% of the time and 100% on of that" type of messages.

MaxAB is in 7.0.xx pretty useless for WCG as you've found out... anything over the official 30 work units per fetch putting you over the MinB level and it's

10777 World Community Grid 1/18/2013 7:44:02 AM Sending scheduler request: To report completed tasks.
10778 World Community Grid 1/18/2013 7:44:02 AM Reporting 1 completed tasks
10779 World Community Grid 1/18/2013 7:44:02 AM Not requesting tasks: don't need
10780 World Community Grid 1/18/2013 7:44:02 AM [sched_op] CPU work request: 0.00 seconds; 0.00 devices

Good "pass over" crunching, well maybe "bridge crunching" is better for the agnostic and other believes part of the population ;>)

MaxAB = "Maximum Additional Work Buffer" fka "Additional Buffer", which worked as a true additional, regardless of whatever. (Not going to enter discussion why that was found no good in multiple respects... visit dev documentation for the reasons).

I'm sorry Sek, I know you're trying to help, but I'm not really following much of this and I know this stuff is second nature to you so I can't expect you to plumb the depths of my ignorance. Plus I'm not really capable of focusing enough to go through the technical docs at the moment. That's a persistent problem but I'm also in the process of taking a break from one of my meds and so I expect it to be even worse for the next few weeks - if recent experience is any indication.

All I know right now is that I've got about 2400 wu's for the octo right now. Since that goes through about 3000k in 24 hours, I'm hoping that if I bump it up a little more, I will make it through the outage without too much of a loss in production. Presumably something similar will be true for the other rigs.