On quad host I have now 4 high priority running SETI tasks that prevent MW tasks (and my ATI GPU) to work.

Why they running in high priority mode?

host details:
% of time BOINC client is running 93.2005 %
While BOINC running, % of time host has an Internet connection 97.9599 %
While BOINC running, % of time work is allowed 99.9867 %
Task duration correction factor 0.21712

Project shares:
2 active projects, MW - 7000, SETI - 10000 .
Cache parameters:
connect every 1 day, 4 additional days of work.
Now most interesting part:
Dealine of high priority tasks is 17 (!)/07, whole week ahead!
There are ~10 MW tasks with deadline 13/07 and estimated time to crunch ~7minutes (!).

So, why BOINC (6.6.36) goes into high priority mode and refuses to crunch MW tasks ?

What another info should I provide to get answer on this question?

[And another question - what can I do to force it leave high priority mode? It started another 4 tasks in high priority again!]

---

The scheduling logic in 6.6.36 is dire. It does not work in so many ways, and there seems to be little enthusiasm for addressing the issues. Your best bet is to suspend the high priority tasks, or the project itself until such time as you think they should run, when you can resume them. I'd recommend suspending the project, otherwise BOINC will keep hassling the SETI servers for replacement work, adding to the download and upload issues that SETI has. In other words do the scheduling manually that BOINC should be doing automatically.

You are not alone.

---

I'm seeing the same thing - my linux server keeps running one project despite cpu sharing rules..

Very annoying.Is this internal Boinc client problem, or is it caused by the project setting priority - and in effect overuling the cpu sharing ??

Frank

---

It happened again.
No active SETI tasks at all, only 4 active Einstein tasks with deadline 1 August (!) and time to complete ~7 hours + active MW project.
All einstein tasks are running in high priority mode totally blocking MW tasks (that use GPU and almost no CPU) from execution.
IMO it means that scheduling logic that governs "high priority" mode transition is completely broken in BOINC 6.6.36 version.
http://boinc.berkeley.edu/trac/ticket/943

---

It happened again.
No active SETI tasks at all, only 4 active Einstein tasks with deadline 1 August (!) and time to complete ~7 hours + active MW project.
All einstein tasks are running in high priority mode totally blocking MW tasks (that use GPU and almost no CPU) from execution.
IMO it means that scheduling logic that governs "high priority" mode transition is completely broken in BOINC 6.6.36 version.
http://boinc.berkeley.edu/trac/ticket/943

And Einstein is one of the well behaving once, try World Community Grid and you get overwhelmed by the number of WU you get.
I always give these projects some time to get new tasks and set them at no new tasks immediately.

But I have never seen the GPU task stop.
I have a system with Einstein in HP and the GPU task running Seti.

...But I have never seen the GPU task stop.
I have a system with Einstein in HP and the GPU task running Seti.

The MilkyWay GPU tasks are not CUDA tasks but (non-intensive?) CPU tasks. Hence the difference to SETI GPU (CUDA) task scheduling.

Gruß,
Gundolf