Hi:

I am here to ask about why version 7.0.64 of BOINC is not trickling my progress, I am a bit afraid that in some point in the computations that it will fail and not be able to recover.

Kiska

What do the model graphics show? What timestep do the models appear to have reached?

(To see the graphics from BOINC Manager, select the model in the Tasks tab and then press the Show graphics button.)

There are 5 tasks with various percentages but when it does communication it requests new tasks which there are none of. One of the hadcm3n is at timestamp 17,223 at 1.657%, one of the hadam3p task is at timestamp 4,224 at 3.096% and another at timestamp 2,304 at 1.799%. The only way I can inhibit a trickle is invoking "no new tasks" command and I do not wish to do that.

You have a 4 core machine.
You have 5 tasks on hand.

Therefore it doesn't need more work, as these models will take it several days in the case of the hadam3 models, and several weeks in the case of the hadcm3 models.

Therefore, if it IS asking for more work, one or more of those models has failed.

********************

Setting the project to NNW just means that BOINC won't ask for more work. It doesn't stop a model producing trickles, or uploading them.
Setting the Network setting to Suspended will stop the uploading of trickles and zips, but not affect the running of models, and hence the production of trickles and zips.

The only real way to inhibit the production of trickles, is to Suspend a model.

********************

Look in the Message window and see if it says somewhere that the scheduler has been invoked with the reason being that it's to send trickle-up message.

... One of the hadcm3n is at timestamp 17,223 at 1.657%, one of the hadam3p task is at timestamp 4,224 at 3.096% and another at timestamp 2,304 at 1.799%. ...

The first trickle for a HADCM3N model is at timestep 25,920, so no trickle will yet have been uploaded. The first trickle for a HADAM3P model is beyond two or four thousand steps, so there will be no trickle there either.

If the timestep is not updating because the model is in some kind of loop, then that would indeed be a problem. Otherwise you just have to wait a bit.

Well now its at timestamp 39,500 and still no trickle in the messages.

There's a trickle on this one: hadcm3n_4hf4_1980_40_008390334_2.