Long WUs high priority?

Thanks for all the detail you have given. It makes it much easier for trying to picture what is happening. One thing you don't seem to have mentioned is your cache setting as controlled by the "connect to network every X days" preference setting. It would have a significant influence if it was too large. Also check that your "switch between projects every ..." is the default value of 60 mins.

Some general comments first. With the values you have listed, EAH should not be getting more work for a long time to come. BOINC should be blocking it for about the next three months or more :). Having said that, I know that suspending and resuming various projects can cause unexpected consequences - particularly if the resource share of the suspended project is large - like that of malaria. That project is supposed to be getting a big share but it doesn't seem to have done all that much work. Do you get regular work from that project and do you ever suspend it? If there's no difficulty getting malaria work it's very hard to understand why BOINC is allowing it to accumulate such a large LTD. Your cache should be awash with malaria WUs :). Another thing to look at is the value for DCF (duration correction factor) that is stored in your state file (client_state.xml) for each project. It would be good to report those five values just to see if any look to be "unusual" :).

If I had your problem, here's what I would do. First off I would check and then reduce if necessary the "connect to network" setting to a very low value like 0.01 days so that no project should attempt to get a whole lot of work. BOINC is far more likely to get itself sorted out quickly if it's not trying to manage a lot of work. There's no indication that your value is too large but you never know. If your value is more than say 1.0 days it is probably too large.

Second, I would resolve not to touch the "suspend" option if at all possible just to force other projects to get extra work. If you suspend the majority of your projects, the remaining ones can think they have full access to all of your cache and can download as if they had 100% share. It's then much harder for BOINC to sort out the resulting mess. Also, I have seen some strange things happen when projects get randomly suspended and resumed.

For those projects like LHC which rarely have work, consider giving them a much higher resource share. Most of the time they wont be able to use it so the other projects will take it back anyway. Then when LHC suddenly has work it can grab a bigger slice because it will have both a higher resource share and a bigger LTD to assist it.

Thirdly, as things are quite out of whack anyway, I would reset all STDs and LTDs to zero just to give BOINC a fresh start. If you decide to change any resource shares, do it first on the website of each project and then update each project in your Boinc Manager to make sure the new resource share has "taken". Then stop BOINC, reset all debts, and restart BOINC.

When BOINC restarts, the low "connect" setting should prevent it from getting too much work from any one project. All projects that have work to give should supply exactly one work unit only. If you have a single EAH work unit, BOINC will probably calculate that it is overcommited and will finish it first. BOINC should then refuse to get any more from EAH for a while.

It may take quite some time for things to settle down but generally speaking you should see BOINC switching regularly between projects that have work. After a week or two when things seem to be going properly, you could consider slowly and carefully increasing your cache in a couple of stages from 0.01 days to something like 0.5 - 0.7 days or so. Allow a couple of days between each increase to get a feel for what is happening and to allow BOINC to cope with the adjustment. Back out the change if there appear to be adverse consequences. Just remember that sudden large changes can cause havoc :).

Good luck and let us know how things turn out.

Ahhh.... That's probably going to turn out to be the source of much wierdness particularly when you start suspending projects. Other projects (eg EAH) can take that as an open invitation to grab a lot of work at that moment. Now that I know that, I would guess that you will probably be able to get much better behaviour just by reducing that 4 day value to 0.5 day. Do that straight away, even before your cache runs down due to "no new tasks" as this will give BOINC time to adjust to the new value. Over the coming days, as project queues get emptied, try allowing new tasks one project at a time to see if that project will then grab a sane amount of new work. Malaria with its 50% share and positive LTD should always be trying to keep its cache full. That's OK as it's got some catching up to do :).

I'm aware of the existence of Boinc Debt Viewer but have no experience with it so I'd be checking any READMEs that come with the package or any FAQs on the website where you got it, just to make sure about that point. You would hope that a button so labelled would be able to do the job without trashing your state file but you are wise to be a little concerned :). If you are going to use that button make sure BOINC is fully shut down first.

EDIT: Actually, you could put that action on hold for the moment to see if the "connect to network" reduction fixes things by itself. As your preference was to give malaria the biggest slice of your time, that's exactly what should happen with the debts the way they are. It would be interesting to see if BOINC is now able to get itself back on track without interfering with the debts :).

The DCFs all look reasonable. The default is 1.0 and the values move up or down slightly over time to correct the project estimate for how long a result should take on your particular computer.