Let me describe the problem again.

I WANT LHC to be 10%.

If I set LHC to 10% it refuses to D/L work.

To force LHC to D/L work, I have to "micromanage" the resource share.

Do you now see the problem?

If the "cannot complete" message overrides the LTD even when LTD is 'critical' then you are totally correct, your machine is not "fast enough" for LHC's short deadlines, at a setting of 10%. The "don't go into EDF mode if possible" rule is overriding the resource share rules. This is not my understanding however of the way BOINC works... if I'm wrong, someone please correct me.

I still believe the problem is in the line "I have to 'micromanage' the resource share". There is no software problem, other than you have not waited long enough for it to do it's job. When LHC's long-term debt is high enough, it WILL download work. As long as you are "forcing" it to download work when YOU think it should, the LTD will never get high enough for it to begin working on it's own. Thus you will be happy that LHC has done 10% of the work "today", but in order to get it to do 10% "tomorrow", you will _again_ have to mess with it.

Get BOINCView, BOINCDV, BOINC Stat Viewer, or whatever is applicable for your OS. Watch the LTD numbers _without_ touching the system. When the LTD (which is in seconds) is 'significantly' greater (10x? I don't know if LTD is adjusted per-share to begin with, in which case it would be 1x, or is adjusted for share only after the fact) than your "connect to every" setting, BOINC will finally _start_ to consider downloading LHC work. Do not be surprised if, given a 10% setting, this does not happen for at least ten DAYS of not changing anything. Once the system has started to work on it's own, it will get LHC work once every few days, giving a long-term result of 10% to LHC. This is the behavior I have seen on a slow Mac laptop with SETI as the "primary" and Predictor as the "backup".

As Ageless said, it will ALSO download work if the "main" project is down or out of work. This will not 'reset' the LTD however, as a project that has no work does not count against your LTD numbers. And in this case, I know that the "idle CPU" rule _does_ override any potential deadline issue. That means that a "down" project can wind up with less than it's assigned resource share, which is only fair; otherwise when a project came back up after a long outage, it would take 100% of everybody's CPU for a significant time while it "caught up".

---