I left a model last night at 99% (my first CPDN model to hopefully finish) and it displayed a computation error this morning. I backed up to an (automated) backup taken just before the completion occurred, and it showed 100% complete.

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=6143188

Claimed credit 27,089.89, Granted credit 39,657.60

CPU time 12151864.770393

Over, Success, Done, 0 (0x0)

Can someone tell me if I have 100% finished my first CPDN model, thank you.

---

Your computers are hidden, so we can not look at the model to see if it did indeed finish.

---

Your computers are hidden, so we can not look at the model to see if it did indeed finish.

Oh Poo :O) How about \"Granted credit 39,657.60\"? Is that what is normally granted for a finished model?

---

From your link direct to the model, it looks like it crashed with a number of negative pressure errors in 2079 - which are fatal (and very bad luck). It was very hot.

Your computers are hidden, so we can not look at the model to see if it did indeed finish.

Oh Poo :O) How about \"Granted credit 39,657.60\"? Is that what is normally granted for a finished model?

For a non-optimised work unit the credit total should be 41,472.00, no matter how long it took you to process it.
For an optimised work unit I am not sure but mine is past this total and still has 13.5% to go.

So if you have 39,657.60 credits then I would say no you tripped just before the line.

I have just had one with 68% completed and over 1700 hours done just go back to 0.095% and Zero hours done (according to the Boinc Manager), I will keep fingers crossed that it sorts itself out as I did not back it up (I have not lost credit on it though).

---

Looks like it trickled for 158 years, so the final granted credit total should be around 40,953. Since it last trickled about 0800 GMT this morning (after the server update), it only had the granted credits through your Sept 12th trickles. The final granted total will show on the next server credit update tonight.

When a cpdn model\'s complete the graph says 1920 - 2080 (BBC models said - 2081 for some unknown reason). So yours missed the last two trickles.

Are you sure that these auto backups presumably done with boinc and workunit running are really reliably restorable? When you restored the backup, did the model crunch at all? Surely it should have crunched for a little while, even a few minutes, until the negative value was created again. If the restored backup showed 100% complete immediately and without crunching, I wonder whether there was something wrong with the backup. Or was your auto backup scheduled to run after the model had already crashed?

If you are backing up frequently and saving at least the 2 most recent backup copies, what about trying to restore the previous backup of this model?

Some incomplete models are generating Over - Success - Done when they really shouldn\'t, but this is a separate issue.

---

Cpdn news

When a cpdn model\'s complete the graph says 1920 - 2080 (BBC models said - 2081 for some unknown reason). So yours missed the last two trickles.

Are you sure that these auto backups presumably done with boinc and workunit running are really reliably restorable? When you restored the backup, did the model crunch at all? Surely it should have crunched for a little while, even a few minutes, until the negative value was created again. If the restored backup showed 100% complete immediately and without crunching, I wonder whether there was something wrong with the backup. Or was your auto backup scheduled to run after the model had already crashed?

If you are backing up frequently and saving at least the 2 most recent backup copies, what about trying to restore the previous backup of this model?

Some incomplete models are generating Over - Success - Done when they really shouldn\'t, but this is a separate issue.

Yes, I was puzzled that a backup taken just before the crash didn\'t run for a while after restore. My backups are reliable enough I think, Norton Ghost 12, and I\'ve gone to a backup 24 hours earlier at 98.5% which is now running with 49 hours to go.

Thanks for your help Mo, and to everyone else. I\'ve upgraded to this model to 5.10.20 and am hoping it won\'t trip again ;O)

---

Some of us think that backups made by a ghosting program with boinc running will sometimes turn out to be restorable but not always. I suspect you\'ve had an example of each.

If you exit boinc before backing up the boinc folder, almost 100% of the backups will be perfect. But if you have a number of computers, I appreciate that the hassle of exiting boinc on all the machines may be too time-consuming to be worth while, particularly if your computers rarely crash models. Easier to make frequent backups with boinc running even if the odd one turns out not to run.

It will be interesting to see whether the model crashes again at the same point. If it does, it really will be impossible climate that the model itself has generated. This is probably the most likely outcome but you never know. Let us know what happens.

---

Cpdn news

Some of us think that backups made by a ghosting program with boinc running will sometimes turn out to be restorable but not always. I suspect you\'ve had an example of each.

If you exit boinc before backing up the boinc folder, almost 100% of the backups will be perfect. But if you have a number of computers, I appreciate that the hassle of exiting boinc on all the machines may be too time-consuming to be worth while, particularly if your computers rarely crash models. Easier to make frequent backups with boinc running even if the odd one turns out not to run.

It will be interesting to see whether the model crashes again at the same point. If it does, it really will be impossible climate that the model itself has generated. This is probably the most likely outcome but you never know. Let us know what happens.

I take your point about the ghosting backup, and the risk v. time available to do a number of backups. But with a number of PCs I\'m crunching I doubt whether I would do manual backups, and with ghost at least I may have half a chance. (I\'m also doing some BBC and SAP as well as SETI, Einstein and Rosetta.)

I re-ran it, after upgrading to the latest version of BOINC, and also running it on a different (better) PC. But it crashed again at the (I think) identical point in May 2079. I took a (ghost) backup 30 minutes or so before it crashed the second time, and I expect I\'ll give it another re-run sometime, just out of curiosity, just in case ;)

---

I wouldn\'t bother trying it again. These negative pressure results mean the model\'s produced climate that\'s impossible in the real world due to that combination of inital parameter values. An unstable computer eg too far overclocked can also produce impossible climate. But if you\'re sure the computer\'s running stably, then it means you got an impossible model. It took a long time to freak out.

If you look at its graph, you\'ll see that it\'s only just an outlier, above the usual upper temperatures.

Still, you have the satisfaction of knowing that you took it as far as it could go.

---

Cpdn news

I wouldn\'t bother trying it again. These negative pressure results mean the model\'s produced climate that\'s impossible in the real world due to that combination of inital parameter values. An unstable computer eg too far overclocked can also produce impossible climate. But if you\'re sure the computer\'s running stably, then it means you got an impossible model. It took a long time to freak out.

If you look at its graph, you\'ll see that it\'s only just an outlier, above the usual upper temperatures.

Still, you have the satisfaction of knowing that you took it as far as it could go.

OK, thanks Mo. I\'ll leave this one be. (I\'m not running any clocked PCs.)

---

In miniature it would make a scary sig.

---

Cpdn news