... and you\'ll lose your computer ID.

Perhaps the following will help:
Before deleting the current install of BOINC etc, save/open client_state.xml
For each Project, locate <rpc_seqno>3934</rpc_seqno>
Write down this number.

When the backup copy has been loaded, repeat the open & locate.
Substitute the saved number. (For each project.)
Now when BOINC talks to the server, the server won\'t know that you\'ve been fiddling.

---

Backups: Here

Ugh. I got two \"2\'s\" in part II. Maybe an admin can fix?

Ugh. I got two \"2\'s\" in part II. Maybe an admin can fix?

Only \'Hide\' and \'Move\' on my screen ...

However, I\'ll still give the method a try on an old backup and see what happens.

addendum to part II.

Windows notepad\'s formatting will give you headaches--literally. Use Win32pad instead

Let me distill my instructions into three sentences:

Delete every trace of the task within the slots and projects/climateprediction.net directories, except for the original zip file.

Set the clock back (after \"<cpu_time_remaining>\") in sched_request_climateprediction.net.xml.

Remove the <active_task> sub-sections in client_state_prev.xml and client_state.xml.

Thanks to MartinNZ for an additional slab iceworld: point #34 on the western map. My hadsm3dhet2 slabs don\'t seem to be generating iceworlds at the same rate as the venerable hadsm3fub. It may just be a small-number effect, with the usual rate asserting itself after a more significant number have been run. We\'ll see.

Thanks to Lockleys for a HADSM3 iceworld: point #35 on the western map.

Since most of the iceworlds end up in the same place it is reasonable to ask whether there\'s any benefit in collecting more of them. One feature of every iceworld analysed in detail so far is that the freeze occurs in the second timestep of a group of six timesteps. I suspect that the \'.cpdn\' file at timestep n actually reports the state of the model in timestep n-1, so the freeze really occurs in the first timestep of a group of six. That will mean something to someone with their hands on the HADSM3 code.

Another diagnostic test is to check whether the freeze occurs at a particular place in the checkpoint cycle. In other words, is an iceworld caused by a problem with HADSM3 file handling and nothing to do with the model itself? There are 144 timesteps in each checkpoint, and therefore 144/6 = 24 possible freeze locations. Here\'s the distribution:

Are the two empty slots significant? By simple probability, the chance of an empty slot with 50 analysed models is ~12%, so we might expect 2-3 empty slots at this point. With 75 iceworlds we might expect 1 empty slot and with 100 iceworlds ~0.3 empty slots. I don\'t honestly think iceworlds are a file problem, but there\'s no harm (though some effort) in excluding the obvious.

The other advantage of more iceworlds is that it might be possible to figure out some parameter dependence. That would be much more physical. I have tried it manually and couldn\'t see any pattern at all - but a more systematic approach might work.

So, keep \'em coming.

Thanks to IanSM for a phase-2 HADSM3 iceworld: point #37 on the ever-popular western map.

Mmmm: what happened to point #36? Ah, a phase-1 slab from MartinNZ - apologies for neglecting that one.

Thanks to IanSM for two more iceworlds: point #38 on the western pile and, like a whale migrating north, a new point off San Francisco ...



... as before, the pattern of \'coastal-ocean\' is maintained.

PS Click on the map for a better quality image.

[Edit: Added another model from MartinNZ. Thanks.]

Thanks to peterfilla for a phase 3 slab iceworld: it\'s point #11 on the updated map.

Thanks to peterfilla for a phase 3 slab iceworld: it\'s point #11 on the updated map.

... and another: point #12 on the updated map.

At last, a Mac iceworld! A development machine of mine has finally produced a fast-processing phase-3 slab. It\'s another west coaster.

That now completes the basic platform coverage. With the exception of Linux/Intel (which doesn\'t have iceworlds, as far as I know) all platforms show the same characteristics at the freeze point: single-point failure in temperature and possibly pressure and possibly precipitation at a coastal ocean location in the second timestep of a group of six. The main difference in subsequent behaviour is that Windows/Intel slows down whereas Windows/AMD, Linux/AMD and Darwin/Intel speed up; phase-1 Darwin/Intel iceworlds crash at the end of the phase, whereas phase-1 iceworlds on other platforms continue as iceworlds until the end of the run; phase 2 iceworlds on all platforms revert to normal processing in phase 3.

Thanks to MartinNZ for point #39 on the west-coast map.

Have aborted a iceworld.
running on a Intel(R) Pentium(R) 4 CPU 3.20GHz [Family 15 Model 4 Stepping 1]
hadsm3fub_jrqw_006446482 resultid=10110946

last show graphics at the day before yesterday shows a normal world temperature.
Today it has changed to a iceworld.

---

Matthias

Have aborted a iceworld.
running on a Intel(R) Pentium(R) 4 CPU 3.20GHz [Family 15 Model 4 Stepping 1]
hadsm3fub_jrqw_006446482 resultid=10110946

last show graphics at the day before yesterday shows a normal world temperature.
Today it has changed to a iceworld.

All three models in that work unit have turned into iceworlds at the same point. One is almost finished, but has been slow-processing since December 2009 - whether deliberately or by accident is impossible to tell. So aborting the model was the right thing to do.

If you have a backup, then you can generate a \'.cpdn\' file that can be processed to find more information about the iceworld. There are instructions earlier in this thread on how to do that.

There\'s a bumper crop this time, both on the western map and the eastern map. Thanks to David Glogau for two iceworlds and MartinNZ for another one. I\'ve been batching mine up and have now added eleven to the pile. The total is now 74 (the maps show 73, plus there\'s Mo\'s outlier near Cyprus).

Courtesy of IanSM and one of mine there are now 76 iceworlds in the big parade!

This task, hadsm3dhet2_k2m8_006613042_8, turned into an iceworld on a Q6600 running Win7_x64, so I aborted it. Sorry, no backup so I can\'t do any recording. It did actually finish, though, on two other machines, one running Linux and the other a Mac.

---

I\'ve reported the Intel + Windows computers in the same WU to Milo who should send an iceworld warning email to their owners.

---

Cpdn news

Is this blue world of any interest?

Name: hadsm3fub_k6pr_006465881
Elapsed: 1774:33:29
Fraction: 89.630%

WindowsXP, Pentium 4