application UK Met Office HADAM3P
created 11 Mar 2009 20:30:21 UTC
name hadam3p_meo5_1994_2_006051207

I send it yesterday and this WU does not gave me any points... someone knows why?

Regards

---

The new server is still catching up on the backlog after last weekend's problems (see news post here). By my very rough reckoning it's still about a day behind in processing trickles.

Your task will be credited when the credit script is next run (and the other 2 you've completed which haven't received their full amount will also be topped up).

---

"The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer

I am crunching 2 models on 2 different pc's. Both report: "Too many total results". Both models do get trickles in what seems a proper interval, but...
Since about a week these trickles don't result in credit in the first model. The second model gets a low number of credits (ca. 94), independent of the number of WU's crunched.
What is wrong? Is there something I can do to solve the problems?

Regards, Roel

Forget about the "Too many total results"; that's just a BOINC thing.
The pages that you should be looking at are: here and here

The first one looks OK, but the 2nd one hasn't sent back any trickles since the 18th. So you won't get any more credit for it until it DOES return more trickles. (No work - No pay.)
Check it in the Tasks tab; what does the Status column say?

I'm not sure what you mean about

The second model gets a low number of credits (ca. 94), independent of the number of WU's crunched.

All models get a fixed number of credits per trickle.
Each different type of model has a different amount of credit, which reflects the amount of work required to get to the trickle point.

---

Backups: Here

Hi Les,

The first paragraph is clear now, thanks.

According to the messages column, the model does send trickle messages. See the last ones below:
24/05/2009 02:19:23 climateprediction.net Sending scheduler request: To send trickle-up message.
24/05/2009 02:19:23 climateprediction.net Not reporting or requesting tasks
24/05/2009 02:20:53 climateprediction.net Scheduler request completed: got 0 new tasks
24/05/2009 06:16:13 climateprediction.net Sending scheduler request: To send trickle-up message.
24/05/2009 06:16:13 climateprediction.net Not reporting or requesting tasks
24/05/2009 06:16:18 climateprediction.net Scheduler request completed: got 0 new tasks
24/05/2009 10:14:06 climateprediction.net Sending scheduler request: To send trickle-up message.
24/05/2009 10:14:06 climateprediction.net Not reporting or requesting tasks
24/05/2009 10:14:11 climateprediction.net Scheduler request completed: got 0 new tasks
24/05/2009 14:10:27 climateprediction.net Sending scheduler request: To send trickle-up message.
24/05/2009 14:10:27 climateprediction.net Not reporting or requesting tasks
24/05/2009 14:10:32 climateprediction.net Scheduler request completed: got 0 new tasks
24/05/2009 18:08:34 climateprediction.net Sending scheduler request: To send trickle-up message.
24/05/2009 18:08:34 climateprediction.net Not reporting or requesting tasks
24/05/2009 18:08:39 climateprediction.net Scheduler request completed: got 0 new tasks
24/05/2009 22:08:20 climateprediction.net Sending scheduler request: To send trickle-up message.
24/05/2009 22:08:20 climateprediction.net Not reporting or requesting tasks
24/05/2009 22:08:25 climateprediction.net Scheduler request completed: got 0 new tasks
25/05/2009 02:10:00 climateprediction.net Sending scheduler request: To send trickle-up message.
25/05/2009 02:10:00 climateprediction.net Not reporting or requesting tasks
25/05/2009 02:11:05 climateprediction.net Scheduler request completed: got 0 new tasks
25/05/2009 06:10:40 climateprediction.net Sending scheduler request: To send trickle-up message.
25/05/2009 06:10:40 climateprediction.net Not reporting or requesting tasks
25/05/2009 06:10:45 climateprediction.net Scheduler request completed: got 0 new tasks
25/05/2009 09:56:12 Not using a proxy
25/05/2009 10:11:21 climateprediction.net Sending scheduler request: To send trickle-up message.
25/05/2009 10:11:21 climateprediction.net Not reporting or requesting tasks
25/05/2009 10:11:26 climateprediction.net Scheduler request completed: got 0 new tasks

The status column says: "Running". The elapsed column keeps counting up in accordance with the time the model is running. As far as I can see everything is going as it should be, with exception of the credits. This makes me fear that there is some hidden problem.

As for the other model, the number of credits did not change when I changes the resources from 1/3 to 1/2 of the pc involved. That is not what I expected, so may be there is a problem too.

Roel

I don't know why those trickles aren't showing on the server. Perhaps someone else will have an idea.

Changing the resources won't change the credits. It will, however, slow down the processing so that it will take a lot longer to get from one trickle to the next, so that will affect the rate at which you get credits, not the amount.

Changing the resources won't change the credits. It will, however, slow down the processing so that it will take a lot longer to get from one trickle to the next, so that will affect the rate at which you get credits, not the amount.

Sorry, I did not make this point clear enough. I ment the credits per day did not change. When the recources go up from 1/3 of the capacity to 1/2, the credits per day should go up by 50%.

Changing the resources won't change the credits. It will, however, slow down the processing so that it will take a lot longer to get from one trickle to the next, so that will affect the rate at which you get credits, not the amount.

Sorry, I did not make this point clear enough. I ment the credits per day did not change. When the recources go up from 1/3 of the capacity to 1/2, the credits per day should go up by 50%.

Could it be that changing the resources will not increase rate of crunching if they were already sufficient to work at fastest.

Just as pouring half a litre from a half litre (or a 750cc) jug to a litre jug will not increase it's volume.

Keith

[Keith Scott wrote:] Could it be that changing the resources will not increase rate of crunching if they were already sufficient to work at fastest.

In the Processor Usage section of the Computing Preferences there are two settings near the end:

• On multiprocessors, use at most: X % of the processors
  
• Use at most: X percent of CPU time
  

Increasing the percentage in the first setting will, as you say, have no effect on the rate at which a single task is processed. A CPDN task runs on one processor at a time: giving it a choice of processors won't speed it up.

Changing the second setting will, however, change the rate at which work is done. People worried about overheating, particularly on laptops, can use this to reduce the heat load.

Perhaps Roel was referring to the first.

:) sounds familiar, i am facing the same problem but on CPDN Beta. mo.v knows the answer.
Regards
Masud.

Both settings are 100%, so that's not the problem.

I have worked on the problems last days and now have a partial solution.
I restarted BOINC on both computers and as that did not make sense I restarted both computers. The result was the same.

Then I upgraded BOINC from 6.6.20 to 6.6.28. That solved the problem on the computer giving 94 credits a day while the number of trickles varied from 1 to 3. It now behaves normal with credits going up and down with the number of trickles.

The model on the other computer still has the same malbehaviour. It generates trickles (according to the messages column). These trickles are not reported in Task details and I don't get any credits.
To see what's on I downloaded another model on the same pc. That model works as it should do. So the problems are restricted the that one model that apart from the trickle problem seems to work as it should do. The graphics look normal too.

It may be coincidence, but the problems started about the time I detached a project (Milkyway) from both computers.

I don't have problems not getting credits anymore for the troublesome model, but as you may understand, I don't trust it anymore. What to do?

Roel

You haven't actually stated which model is not showing trickles.

So just to be clear, is it hadsm3mh_kqi2_006023614_6, on this Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz computer?

And are there two models running on that computer, the other one of which is hadam3p_n6su_1964_2_006157304_1?

edit
I've just had another thought:
When a model rewinds, (for numerous reasons), the trickles that now get sent back, are the same as those that have already been sent, so they don't show up on the server, and of course, you won't get credit for the duplicates.
For the model hadsm3mh_kqi2_006023614_6, the server shows that the last trickle received is timestep 43,208, in phase 3.
What timestep and phase is showing on the graph display for that model?

---

Backups: Here

It's the first model ending 14_6.

Your thought was right to the point. The model now in phase 2, the timestep about 193000.

Why should a model rewind so much?

Roel

Hi Roel

The problem model's web page is here.

It had sent a few trickles from phase 3 but is now repeating phase 2. Fortunately the computer is nice and fast so the model will soon be back to where is was before the rewind.

I think you probably did something on the computer that the model didn't like and this caused it to rewind to the beginning of phase 2. Is is possible that you closed down the computer but didn't completely exit from BOINC first? Doing this can cause a rewind or a model crash. If you have BOINC installed as a service remember that exiting is a two-stage process.

BOINC Manager >Advanced view >Advanced >Shut down connected client >OK >Cancel. Then BOINC Manager >File >Exit.

When the model has completely finished and uploaded its last file to the server, its stderr out messages on its web page will become available. These messages may explain the cause of the rewind.

---

Cpdn news

Today the model is back to the point where it rewinded in the 18th of May. So, as expected, it started to report trickles again.
From the CPU-time, which has doubled since May the 18th, it is claer that the model started all over again from the very beginning. I must have ill-treated it very severely to cause such a thing.

Roel

It is not a lucky model; now it has turned into an iceworld. I am going to abort it.

Well, you gave it the benefit of the doubt, and you can't do much more than that. Some models are just going to fail anyway, because of the particular starting values used. That's what the modelling is about.

Hope that you have better luck with the next one.