Here it is, ten minutes after I snoozed BOINC, and Climate Prediction is still taking up a whole core.

I tried "snooze" because the Climate Prediction project wasn't suspending computation according to my preferences (i.e., when the computer is in use). My other projects don't seem to exhibit this issue.

The Activity>Suspend menu item doesn't seem to have an effect, either.

In the BOINC Manager 'Tasks' tab (7.0.31), everything is marked with a status of "Suspected - user request", but the Mac activity monitor still shows hadam3p_pnw_6.09_i686-apple-darwin as consuming 100% of the CPU.

Any hints? The usability of my Mac (OS X v10.8.3) is suffering mightily.

Thanks,

-Bill

The only thing that comes to mind is that you're talking about the server settings in your post, and that you also have the local prefs set in your manager's menu. (To something different.)

Just in case: it's easy to create these local prefs accidentally, simply by looking at them and then clicking OK to exit. The only safe ways to exit are by clicking Cancel, or the big red X.

Actually there IS something else.

If you're using the server settings, go to your account page and check that the computer in question is set to use the same "venue" for which the Computing prefs are set.

Hmm.

I made my account preferences consistent (and deleted all venue settings).
I deleted my local preferences (which had differed from the online preferences).
BTW, I've since upgraded to the April 25, 2013 build of BOINC.
But I'm still seeing the same issue with hadam3p_pnw_6.09_i686-apple-darwin.

As a test, I set the BOINC Manager "Activity" menu to "Run Always", and I saw multiple BOINC tasks pick up until all eight cores were saturated (only one task running was climate prediction). Once that got going for a few minutes, I clicked "snooze", and within ten seconds (with updates every five), the Mac's Activity Monitor showed all the tasks taking zero CPU, except hadam3p_pnw_6.09_i686-apple-darwin still pulling two cores worth.

In the BOINC Manager "Projects" tab, I can select climateprediction and click "suspend".
When I look up "Properties" for climateprediction, it states "Suspended via GUI = yes".
Yet I still see the hadam3p_pnw_6.09_i686-apple-darwin keeping two cores busy (again, still true after ten minutes).

Is there a glitch on the Mac platform? Or has my unit gotten into a weird mode? (There's one task at 100.000%... might be stuck...).

The CPU consumption at two cores (of eight) seems to be completely ignoring my suspend commands, regardless of the UI used. It also seems to ignore the 'suspend while in use' preference. Project 'Properties' for climateprediction shows 'suspended = yes'. Likewise, the task status in the task list, does too, with a further proper differentiation between the causes 'user request' and 'computer is in use'.


Thanks,

-BIll

NO climate model at present uses more than a single core.

However, each task DOES have 3 programs "in the system": one runs the modelling, one is used to "gather up the data" to create a single zip file to return to the project, and I've forgotten what the other one's for.
But only one will be running at any one time.

---

Backups: Here

When was BOINC Manager upgraded?

The reason for asking is that upgrading BOINC on a Mac while there are still some models downloaded usually crashes those models. It will be interesting to see whether any trickles have been submitted since the upgrade.

(This won't, of course, be an explanation for the original problem which presumably pre-dates the upgrade.)

I now have version 7.0.65, which I believe I had downloaded and installed about 25-APR-2013.

Les,

It seems that something had hanged: A task showing a persistent 100.000% completion was burning a core full time, no matter what the settings. I snoozed BOINC, killed the still-running process from Activity Monitor, quit BOINC Manager, then relaunched it.

In BOINC Manager, the task (at 100% completion) appeared briefly, then disappeared. Maybe submitted, maybe just lost, I'm not sure.

Its hard to say how long that had been the case, since I was only paying attention to CPU load since around the time I started this thread. Plausibly, it's been going on since the last time something updated.

I'll keep a casual eye out for a recurrence, but for now... problem gone.

Thanks for thinking on the problem...

-Bill

A week or two ago another CPDN member reported the same thing about a model completed to 100% that continued apparently processing using a core until the user aborted it. In both cases all the model's files will have been successfully uploaded so even if it's classified as aborted on its web page it will be complete and suitable for use by the researchers.

I can't remember whether the previous case happened on a Mac.

My guess is that these models got into a loop right near the end and continued processing the part between the last two checkpoints. This sort of loop can continue for quite a long time, though not indefinitely.

---

Cpdn news

In this case, the last Timestep received was 518,400, out of 1,036,800. This is the halfway point, so that model failed at 50%, a common place to fail, as has been posted often here.

The "100%" was because, when BOINC gets a message that these models have stopped due to any of many reasons at a 25% point, it thinks that the model has completed, and so displays that 100% message.
It's necessary to look at what the model says about where it's up to, not what BOINC thinks.

---

Backups: Here

Bill, take Les's word for it, not mine because I didn't look at the model's web page whereas he did. Ignore what I said.

---

Cpdn news