Hi there.

My leading CPND model is at about 80% but keeps on dying with an \'unrecoverable error\' I have restored from a backup several times but the same death occurs each time.

Could someone check this model to see if it has gone \'outside the model parameters\' or whether there is some other more fixable problem. I hate not getting one of my models to the end, but suspect that this might well be my first.

Model details and messages as follows:-

11/05/2007 13:52:30||Starting BOINC client version 5.4.11 for windows_intelx86
11/05/2007 13:52:30||libcurl/7.15.3 OpenSSL/0.9.8a zlib/1.2.3
11/05/2007 13:52:30||Data directory: C:\\Program Files\\BOINC
11/05/2007 13:52:30||Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 4400+
11/05/2007 13:52:30||Memory: 2.00 GB physical, 3.85 GB virtual
11/05/2007 13:52:30||Disk: 275.08 GB total, 5.18 GB free
11/05/2007 13:52:31|climateprediction.net|URL: http://climateprediction.net/; Computer ID: 652693; location: ; project prefs: default
11/05/2007 13:52:31||General prefs: from climateprediction.net (last modified 2007-01-10 10:51:56)
11/05/2007 13:52:31||General prefs: using your defaults
11/05/2007 13:52:31||Local control only allowed
11/05/2007 13:52:31||Listening on port 31416
11/05/2007 13:52:31|climateprediction.net|Deferring task hadcm3pbb_che1_05856484_0
11/05/2007 13:52:31|climateprediction.net|Deferring task hadcm3ohe_0t64_05686142_1
11/05/2007 13:52:31|climateprediction.net|Deferring task hadcm3pbb_c46i_05839365_1
11/05/2007 13:52:31||Suspending network activity - user request
11/05/2007 13:52:53||Rescheduling CPU: result suspended, resumed or aborted by user
11/05/2007 13:52:54|climateprediction.net|Restarting task hadcm3pbb_che1_05856484_0 using hadcm3 version 515
11/05/2007 23:56:39||Rescheduling CPU: application exited
11/05/2007 23:56:39|climateprediction.net|Computation for task hadcm3pbb_che1_05856484_0 finished
11/05/2007 23:56:40|climateprediction.net|Unrecoverable error for result hadcm3pbb_che1_05856484_0 (<file_xfer_error> <file_name>hadcm3pbb_che1_05856484_0_14.zip</file_name> <error_code>-161</error_code></file_xfer_error><file_xfer_error> <file_name>hadcm3pbb_che1_05856484_0_15.zip</file_name> <error_code>-161</error_code></file_xfer_error><file_xfer_error> <file_name>hadcm3pbb_che1_05856484_0_16.zip</file_name> <error_code>-161</error_code></file_xfer_error>)
11/05/2007 23:56:40|climateprediction.net|Deferring scheduler requests for 1 minutes and 0 seconds


Advice would be most welcome. I have not allowed the model to attach to the project yet to report its failure.

Cheers.

Pete McCann

These messages tell us nothing useful. The 161 errors just mean that there were no final zip files to upload, which is obvious, seeing that the model HASN\'T finished, which is what BOINC thinks.

The useful messages which should contain the REAL error code and/or \"Negative ...\", won\'t appear until you allow an upload.
The Negative Theta / Negative pressure part should be in the many files somewhere, but I don\'t know which one.
Someone who does know should be along eventually, so keep watching this thread.

Basically, if the model keeps crashing at or about the same year/month/day/hour, then it\'s probably never going to get any further.

Hi,

The model may have reached as far as it is going to go - if it\'s always stopping at the same point, and the error text contains \'NEGATIVE PRESSURE DETECTED\' (or NEGATIVE THETA DETECTED).

If present, the NEGATIVE ... indicates that the model physics have become unrealistic. It is one of the goals of the project to find which combinations of starting parameters are viable, and which are not. It is generally not worth restoring from a backup if you see these messages (they\'ll only stop again at the same point).

Since it hasn\'t reported yet, and hence we can\'t see the error text, we can\'t tell if this is the case or not.

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Since it hasn\'t reported yet, and hence we can\'t see the error text, we can\'t tell if this is the case or not.

Mike.

I will allow the model to upload, as it is now showing as 100% completed in my \'tasks\' tab. Maybe you could have a look for me once you have acess to it and tell me whether it is worth persisting or to call it a day on this one.

Thanks again for your time.

Cheers.

Pete McCann

Model crashed: umshell1.f: P_TH_ADJ : NEGATIVE PRESSURE VALUE CREATED. GA

Model crashed: umshell1.f: P_TH_ADJ : NEGATIVE PRESSURE VALUE CREATED. GA

Model crashed: umshell1.f: P_TH_ADJ : NEGATIVE PRESSURE VALUE CREATED. GA

Model crashed: umshell1.f: P_TH_ADJ : NEGATIVE PRESSURE VALUE CREATED. GA
Fatal crash! :-(

Looks like it has reached as far as it can on that particular set of parameters. On the more positive side, it\'s got past the 2050 point where models will get counted as \'completed\' on the front-page of the website, so congratulations :-)

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Thanks Mike.

Ah well. Hopefully knowing what doesn\'t work is just as valuable as knowing what does. So will this model count as completed for Team Stats purposes and for me personally, even though it hasn\'t made it to 2080?

Can you tell from the data what the \'unrealistic\' element in this particular model was?

On with my other 14 models I suppose then!

Cheers.

Pete

I think it must be air pressure which in the real world must always have a positive value. I imagine it could only ever be negative if the Earth\'s entire atmosphere were sucked away by some outside force? Better perhaps not to explore this scenario.

I don\'t know what your team stats count as a completed model, but the figures I quote from Oxford in the News thread include, as Mike said, all models that reach 2050 or beyond. It\'s obviously better for the researchers to reach 2080 if your model parameters allow it. A few other crunchers will envy your 14 other chances!

Every cruncher who takes a model as far as its parameters allow it to go has scored a personal success.

---

Cpdn news

The team England stats count 2080 as a completed model, as far as I know Scotland and Wales use the same criteria. If CPDN report anything past 2050 as complete, this leaves us singing from different hymn sheets.
If the rules of the game are to be changed, they need to change for all teams.
Although point scoring isn\'t the aim of CPDN or any other project, it does help produce a lot of research.
If it\'s a goal once the balls in the penalty box we\'ll have to rescore all the games.

Back in January I said in the news thread that Carl had started counting 2050+ as \'finished\' for the purposes of quoting completed models on the cpdn website front page. Here\'s the news item:

http://www.climateprediction.net/board/viewtopic.php?t=5927&postdays=0&postorder=asc&start=0

As far as I\'m aware, all the teams that are counting completed models have only ever counted 2080 as a goal. Could I respectfully suggest that the best idea is probably to keep it that way. Otherwise you could all spend weeks working out which extra models count and which don\'t. The BBC teams would probably also need to agree, which would complicate matters. I don\'t think it matters one bit that the cpdn front page uses a different criterion.

The teams all have a good easy-to-understand system and it\'s been the same from the start. I\'d stick to it.

---

Cpdn news

Back in January I said in the news thread that Carl had started counting 2050+ as \'finished\' for the purposes of quoting completed models on the cpdn website front page. Here\'s the news item:

http://www.climateprediction.net/board/viewtopic.php?t=5927&postdays=0&postorder=asc&start=0

As far as I\'m aware, all the teams that are counting completed models have only ever counted 2080 as a goal. Could I respectfully suggest that the best idea is probably to keep it that way. Otherwise you could all spend weeks working out which extra models count and which don\'t. The BBC teams would probably also need to agree, which would complicate matters. I don\'t think it matters one bit that the cpdn front page uses a different criterion.

The teams all have a good easy-to-understand system and it\'s been the same from the start. I\'d stick to it.

It\'s easy for you to say that mo.v. But you, the moderators, have been telling people that, despite their models crashing at 2050+ they can consider their models are complete, thank you very much, please start another one in CPDN.

I think perhaps it is you that should say that, despite the fact that you have indicated a 2050+ model is complete, it is not, fact, complete for any criteria apart from what is being displayed in the CPDN front page. Which has got nothing to do with whether a model is complete or not (ie, 2080, zips uploaded, received, update acknowledged, credits allocated, goal by ref and linesman confirmed.)

---

So will this model count as completed for Team Stats purposes and for me personally, even though it hasn\'t made it to 2080?

Easy fellas. I wasn\'t trying to start a scrap here. The 2050 date was news to me for a \'completed\' model. If it makes sense for the CPND headline stats then fine. Sounds like a PR job to me! I personally feel a completed model should be the whole 160 year run for individual and team purposes, that\'s despite being rather miffed that one of my models has been unable to complete it\'s full run, due to it becoming unrealistic and outside normal parameters and no fault of mine. But hey, that\'s life, and hopefully the infomation from the unfinished model is just as useful. Presumably one of the points of the whole exercise is to edge slowly towards a better global climate model, and finding out what works and what does not helps in this aim. I\'ve got my positive and optomistic hat on here!

Onwards everyone!

Pete McCann

Hi crunchers!

I think Peter\'s right - I guess that when Carl said he was going to include 2050+ models in the front page figures, he was probably in a PR mood.

Ice! I challenge you (in the nicest possible way) to find a single post by a mod where any of us gave the impression that a 2050+ crashed model could just be abandoned because it had reached its target. If it couldn\'t be saved, we\'ve been consoling the cruncher by telling them that the model would be very useful to the researchers because it had got so far, and it would be counted in the front-page figure. In every case where such a model could be saved, we\'ve been helping members to save it. My news item on Jan 9 announcing the change said

Carl is now counting models that have reached 2050 as completed because they don\'t need to have reached 2080 to be really useful to the Oxford researchers. Reaching 2080 is still of course the ideal.

My news item of 3 March said

It\'s still preferable to get models all the way to 2080 if possible, so keep backing the whole BBC or boinc folder up! How-to instructions in the READMEs.

And on 31 March I said in the News

If your model has reached 2050, it\'s still much more useful to the researchers if you can keep going - the real finishing line is still 2080!


Which is what all the teams have understood all along. What we all want are real goals in the net. But if we played hard and just missed (especially if it was through no fault of our own), we still deserve to be told how well we played and how much we contributed to the collective success.

---

Cpdn news

I think Peter\'s right - I guess that when Carl said he was going to include 2050+ models in the front page figures, he was probably in a PR mood.

Ice! I challenge you (in the nicest possible way) to find a single post by a mod where any of us gave the impression that a 2050+ crashed model could just be abandoned because it had reached its target.

I think you may be missing the point here by looking at this through your CPDN/PR angle.

This is like a Blair/Bush spin. It doesn\'t matter what you think you said and how well you think you can challenge anyone on what you actually said.

The fact is that you have given an impression that models at 2050+ are complete. I expect some are confused about this.

The fact is that if someone decides to start up a \"models complete\" competition in BBC or CPDN, given the spin in place, anyone can stop crunching at 2050 and quicken their model completion by 30 years. No matter what you said or think you said.

Since you and the moderators have said much on this subject, in so many \"official\" and individual statements of encouragement which clearly indicate that a model is complete at 2050+, including the PR on the CPDN front page, I think you should make a definition of what is a \"complete model\" clear to both BBC and CPDN and reconcile this with Carl\'s PR on the CPDN web front page.

If Carl thinks a model is complete at 2050+, why do we need to carry on to 2080?

It\'s a rhetorical question, but illustrates the lunacy of the situation and why Blair/Bush type spin should never be welcomed anywhere.

---

Ice, please think about the physics and science of it.

By the time the climate gets to 2080, there are many assumptions in the model which become somewhat tenuous. How much CO2 will humans be producing? How many humans will there be? Will carbon sequestration be in use? Will fusion be practical?

None of these things are known, therefore the far modelling involves more guesses than the earlier modelling.

Carl has called the 2050-2080 period \'the icing on the cake\' from the scientist\'s viewpoint. They\'re most interested in 1920-2050, but despite that, 2050-2080 still has some use.

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Hi ICE, my fellow Team England (BOINC) team mate.

I think you might be reading a bit too much into all this. A model that makes it past 2050 will have completed at least 40 years of the future predictive phase of the model and will still give a good indicationas to where that model trend was heading. The 2050 to 2080 period is also going to be the most unreallistic part of any model run as this is where you are furthest away from the starting point and any errors or unrealistic variables are multipled in their consequences. Short term predictions are always more accurate than long term ones. From a science point of view I can see that the 2000 - 2050 or even the 2000 - 2030 periods are of greater reliability / worth than the 2050 - 2080 period predictions. That\'s not to say that the last 30 years are a waste of time as they will show the the models trend over a greater period of time. I can see why Carl opted to use models that have reached 2050 and prematurely crashed as complete, as it will be good for the projects public image to have more completed models, and also still of significant scientific value. My model that just crashed in the 2050\'s will still show a climate trend, but it just happens to have gone outside the predetermined limits set for all the models.

I think that everyone is singing from the same hymn book as everyone wants their models to do the full 160 year run and finish at 2080. The stats and teams are a bit of fun, to make the tough slog of crunching these models more enjoyable. I\'m sure we are all passionate aboout the aims of running these models and that is why we are all doing it in the first place. The numbers game of completed models, RAC, credit, and league tables is all incidental to the main aim of advancing the science. Well it is for me anyway! I done think that anyone is seriously suggesting \'moving the goal posts\' (to continue the football theme) and changing the basis of counting completed models. I\'m sure we can all cope with two different measure of \'completed\' to be used for two different purposes. \'Proper Complete\' vs \'PR complete\'! Sure, I would have much rather had another proper \'completed\' model on my stats, especially given how long each model takes to run, but this is science and life, and \'sh*t happens\' It is a bit disappointing for me individually and for the team not to notch up another success, but at least this is just a hobby and not life and death. (Well not for a few more years yet anyway!)

I\'ll just have to build some more computers that\'s all. I\'ve got a nice little P4 840 extreme edition on an Asus P5B deluxe with 2 GB Corsair PC6400 ram on the bench as we speak. UUmmmm. Tasty!

Cheers.

Pete McCann

Hi Mike.

I think you got there just before me, and said the same thing in far fewer words!

Cheers.

Pete McCann

My own BBC model crashed in 2064 after being moved to a new and much faster computer. The increase in speed caused an unusual sort of CPU timeout. I\'ve taken the trouble to restore quite an old backup (my fault there!) and edit the model\'s xml files to enable it to run right to the end.

So while 2050 is very good for the researchers, 2080 is perfect.

I hope Milo will soon be making an announcement about how he\'s making a selection of our model results available for researchers worldwide. I\'m pretty sure he only uses models that reached 2080.

---

Cpdn news

My own BBC model crashed in 2064 after being moved to a new and much faster computer. The increase in speed caused an unusual sort of CPU timeout. I\'ve taken the trouble to restore quite an old backup (my fault there!) and edit the model\'s xml files to enable it to run right to the end.

Sounds a bit teckie to me! I\'ll let the \'dead dog lie\'.

Pete

Peter, yours couldn\'t be saved as it had started to produce results impossible in the real world. When this happens the models automatically crash, which is a good thing.

---

Cpdn news

Ice, please think about the physics and science of it.

My question was NOT about the science of it. It is about teams bragging that they have completed more models than other teams, when the definition of a complete model is whatever you want it to be.

---