climateprediction.net home page
Iceworld Appeal

Iceworld Appeal

Message boards : Number crunching : Iceworld Appeal
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 7 · Next

AuthorMessage
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 37861 - Posted: 21 Aug 2009, 11:24:16 UTC

As Mo says, I\'m collecting iceworlds in an attempt to gather evidence about what their cause might be: good progress is being made.

What I need is the \'playback\' file at the moment the iceworld freezes. This may seem like a hard thing to do, but it\'s not really. Playback files are generated when graphics recording is turned on (by pressing Control-Q when the graphics are running). The files accumulate in the \'tmp\' folder for the model - e.g. \'T:\\Data\\BOINC\\projects\\climateprediction.net\\hadsm3mh_ktk0_006316540\\tmp\' for a Windows model I\'m recording at the moment - but your model will be in a slightly different place. The files have the extension \'.cpdn\' (e.g. \'hadsm3mh_ktk0_006316540.0021574.cpdn\') and are about 120 kB, so the amount of disk space used rises rapidly. When the iceworld freezes, the size of the file drops to about 8 kB (and 3 kB eventually). The file I need is the last full-size \'.cpdn\' playback file before the drop.

To do this you would need to make a backup of your current working installation, then restore a backup from before the freeze. Turn network activity off and re-run the iceworld until it freezes. Send me the single \'.cpdn\' file and then delete the iceworld installation and restore the working installation. Then carry on as normal.

If you would rather just get on with some more processing, then just ignore this post! I quite understand.

Thanks,

Iain

PS By the way, the graphics do not need to be left running when the recording has started. However, make sure you\'ve got enough disk space: one phase is 10802 * 24 * 120 kB = ~30 GB, though you shouldn\'t have to run it that long.
ID: 37861 · Report as offensive     Reply Quote
Profile mo.v
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 2363
Credit: 14,611,758
RAC: 0
Message 37862 - Posted: 21 Aug 2009, 14:12:26 UTC
Last modified: 21 Aug 2009, 14:14:41 UTC

Thanks for unhiding your computer. Is this the iceworld model?

What a lot of models you\'ve processed for CPDN.
Cpdn news
ID: 37862 · Report as offensive     Reply Quote
old_user85070

Send message
Joined: 27 Jun 05
Posts: 6
Credit: 1,733,861
RAC: 0
Message 37864 - Posted: 21 Aug 2009, 14:42:54 UTC - in response to Message 37862.  

Thanks for unhiding your computer. Is this the iceworld model?

What a lot of models you\'ve processed for CPDN.


Yes that is it. I have kept backups every day to ensure I don\'t corrupt too many. Spybot destroyed a few when I was in graphics mode. I fully support your aims and we can only change things if we give enough details to sway the doubters.

Dibb
ID: 37864 · Report as offensive     Reply Quote
old_user85070

Send message
Joined: 27 Jun 05
Posts: 6
Credit: 1,733,861
RAC: 0
Message 37866 - Posted: 21 Aug 2009, 23:23:24 UTC - in response to Message 37861.  

As Mo says, I\'m collecting iceworlds in an attempt to gather evidence about what their cause might be: good progress is being made.

What I need is the \'playback\' file at the moment the iceworld freezes. This may seem like a hard thing to do, but it\'s not really. Playback files are generated when graphics recording is turned on (by pressing Control-Q when the graphics are running). The files accumulate in the \'tmp\' folder for the model - e.g. \'T:\\Data\\BOINC\\projects\\climateprediction.net\\hadsm3mh_ktk0_006316540\\tmp\' for a Windows model I\'m recording at the moment - but your model will be in a slightly different place. The files have the extension \'.cpdn\' (e.g. \'hadsm3mh_ktk0_006316540.0021574.cpdn\') and are about 120 kB, so the amount of disk space used rises rapidly. When the iceworld freezes, the size of the file drops to about 8 kB (and 3 kB eventually). The file I need is the last full-size \'.cpdn\' playback file before the drop.

To do this you would need to make a backup of your current working installation, then restore a backup from before the freeze. Turn network activity off and re-run the iceworld until it freezes. Send me the single \'.cpdn\' file and then delete the iceworld installation and restore the working installation. Then carry on as normal.

If you would rather just get on with some more processing, then just ignore this post! I quite understand.

Thanks,

Iain

PS By the way, the graphics do not need to be left running when the recording has started. However, make sure you\'ve got enough disk space: one phase is 10802 * 24 * 120 kB = ~30 GB, though you shouldn\'t have to run it that long.

Iain
I have run (4adsm3fub_kgbp_005985944_7) till it started to turn blue. But I am having trouble finding the file you want. Can you tell me where it is?
Dibb
ID: 37866 · Report as offensive     Reply Quote
old_user85070

Send message
Joined: 27 Jun 05
Posts: 6
Credit: 1,733,861
RAC: 0
Message 37867 - Posted: 22 Aug 2009, 1:09:59 UTC - in response to Message 37862.  

Thanks for unhiding your computer. Is this the iceworld model?

What a lot of models you\'ve processed for CPDN.

Mo
Which file does Iain want? he said \".cpdn\" file but I can\'t see one with that attribute
Dibb

To do this you would need to make a backup of your current working installation, then restore a backup from before the freeze. Turn network activity off and re-run the iceworld until it freezes. Send me the single \'.cpdn\' file and then delete the iceworld installation and restore the working installation. Then carry on as normal
ID: 37867 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 37871 - Posted: 22 Aug 2009, 9:30:03 UTC
Last modified: 22 Aug 2009, 9:49:48 UTC

Dibb,

There are various ways to install BOINC and the files can end up in different places depending on the choices made during installation. Your machine is running Windows XP, so the \'tmp\' folder in which the playback files are stored will be something like:

C:\\Documents and Settings\\All Users\\Application Data\\BOINC\\projects\\climateprediction.net\\hadsm3fub_kgbp_005985944_7\\tmp

If graphics recording was turned on before the iceworld froze, then the \'tmp\' folder should now be filled with thousands of files all ending \'.cpdn\'. The earlier files will be about 120 kB and the later files about 8 kB.

The file I need is the last full-size one (i.e. 100-120 kB) before the file size drops.

Once the file has been identified, I can send a private message on this board with an e-mail address to which the file can be sent.

Thanks for re-running the iceworld. That\'s very good of you.

Iain

[Edit: the BOINC FAQ lists a number of places where BOINC may have installed its data files: The Big BOINC 6 Answer Thread.]
ID: 37871 · Report as offensive     Reply Quote
old_user85070

Send message
Joined: 27 Jun 05
Posts: 6
Credit: 1,733,861
RAC: 0
Message 37872 - Posted: 22 Aug 2009, 14:51:28 UTC - in response to Message 37871.  

Dibb,

There are various ways to install BOINC and the files can end up in different places depending on the choices made during installation. Your machine is running Windows XP, so the \'tmp\' folder in which the playback files are stored will be something like:

C:\\Documents and Settings\\All Users\\Application Data\\BOINC\\projects\\climateprediction.net\\hadsm3fub_kgbp_005985944_7\\tmp

If graphics recording was turned on before the iceworld froze, then the \'tmp\' folder should now be filled with thousands of files all ending \'.cpdn\'. The earlier files will be about 120 kB and the later files about 8 kB.

The file I need is the last full-size one (i.e. 100-120 kB) before the file size drops.

Once the file has been identified, I can send a private message on this board with an e-mail address to which the file can be sent.

Thanks for re-running the iceworld. That\'s very good of you.

Iain

[Edit: the BOINC FAQ lists a number of places where BOINC may have installed its data files: The Big BOINC 6 Answer Thread.]

IAIN
I have the file identified now. Where do I send it
Dibb
ID: 37872 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 37873 - Posted: 22 Aug 2009, 16:30:32 UTC - in response to Message 37872.  

I have the file identified now. Where do I send it

I have just sent you my e-mail address in a private message on this message board. You should be able to read the message using the \'Inbox\' link at the top-right of the page.

(If that doesn\'t work I\'ll create a public address on Hotmail or the like and post it here.)

Thanks for your efforts.

Iain
ID: 37873 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 37880 - Posted: 23 Aug 2009, 13:10:49 UTC

Dibb,

That was a text-book iceworld: thank you for taking the trouble to find it and send it over.

The slab model went out of bounds in temperature and pressure on the west coast of North America. Here is an image of the models analysed so far: your model is the second of the north-western pair. All ice worlds so far have crashed in the ocean adjacent to land. Two other models show the same behaviour in the Mediterranean. Given all the possible coastal locations for a crash, presumably the limited latitude range is significant - however, that\'s a matter for the project physicists.



Green tinted rectangles/blobs are model land, blue tinted are model ocean.

Iain
ID: 37880 · Report as offensive     Reply Quote
old_user85070

Send message
Joined: 27 Jun 05
Posts: 6
Credit: 1,733,861
RAC: 0
Message 37891 - Posted: 24 Aug 2009, 9:11:07 UTC - in response to Message 37882.  

These graphics are amazing.

I think we should get these very interesting posts about iceworlds and Iain\'s graphics into a thread of their own because the iceworld phenomenon is not connected to the \'client detached\' redesignation.

If Dibb and Iain don\'t object I\'ll split this thread after about 12 hours. A relevant title would make the iceworld posts more easily searchable.


Everything has been sorted out between us now. My detached WU still show as new even though I have aborted the Iceworld and finished the other Slab. Do you have any follow up as to when it will show as completed?
Dibb
ID: 37891 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 37992 - Posted: 13 Sep 2009, 13:55:59 UTC

Dibb has kindly sent me another slab iceworld, and the previous graphic has been updated accordingly.

This brings the total number of iceworlds processed in this way to 19: 9 slabs and 10 mid-holocene, all Windows, split 18 Intel and 1 AMD. I would be particularly interested in a Darwin or Linux/AMD fast-processing iceworld if anyone\'s got one (i.e. prepared to re-run a backup of one, or spotted an iceworld ahead in the work unit and prepared to switch on the graphics recording and send over the playback file at the crash point).

(So far as I can tell Linux/Intel does not have iceworlds.)

Thanks.
ID: 37992 · Report as offensive     Reply Quote
wateroakley

Send message
Joined: 6 Aug 04
Posts: 185
Credit: 27,083,655
RAC: 6,161
Message 38039 - Posted: 27 Sep 2009, 13:51:19 UTC

Hi Iain, I have another iceworld today H S E. Just copying yesterday\'s backup to re-run it. I\'ll send you the requested files when it crashes.

ID: 38039 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 38043 - Posted: 28 Sep 2009, 10:12:38 UTC - in response to Message 38039.  

Hi Iain, I have another iceworld today H S E. Just copying yesterday\'s backup to re-run it. I\'ll send you the requested files when it crashes.

Thanks. I guess we\'re on the tail end of the the mid-holocene models, but they behave exactly the same as the slabs from what I can tell. Originally, I thought MHs produced more iceworlds, but that was just because I\'d had a really bad run - now I\'ve processed more the rate seems a bit less, if anything (MH phases 2 and 3 have relatively fewer freezes). So I suspect that switching back to slabs will increase the hit rate somewhat.
ID: 38043 · Report as offensive     Reply Quote
wateroakley

Send message
Joined: 6 Aug 04
Posts: 185
Credit: 27,083,655
RAC: 6,161
Message 38045 - Posted: 28 Sep 2009, 15:02:07 UTC
Last modified: 28 Sep 2009, 15:04:22 UTC

Iain, the backup server running your task failed overnight (a dodgy second HDD). The task has continued OK after disconnecting the offending HDD. It\'ll be a couple of days to get to the crash point.

The WU has 8 companion tasks. It looks like task 9363338 H S E has crashed at/around the same point in phase 4. Task 9363340 is on Intel/Linux but not progressed as far yet. The rest are Intel with a mixture of MS variants.
ID: 38045 · Report as offensive     Reply Quote
wateroakley

Send message
Joined: 6 Aug 04
Posts: 185
Credit: 27,083,655
RAC: 6,161
Message 38046 - Posted: 30 Sep 2009, 6:51:06 UTC
Last modified: 30 Sep 2009, 6:53:01 UTC

Iain, the files are on their way to you.

Edit: typo.
ID: 38046 · Report as offensive     Reply Quote
Profile JIM

Send message
Joined: 31 Dec 07
Posts: 1152
Credit: 22,053,321
RAC: 4,417
Message 38047 - Posted: 30 Sep 2009, 23:41:13 UTC
Last modified: 30 Sep 2009, 23:42:41 UTC

Hi Ian:

I have an early, slow processing ice world for you. WU hadsm3fub_jwtx_006408687_1 froze at 14.96% at 26/8/1817 at 10:30. All clouds have disappeared and the temp graphic is uniformly blue. I am running Windows 7RC on an Intel core2duo 1.5 GHz processor with 2 GB’s of RAM. The machine was otherwise idle at the time.

I restored from a backup, and ran it to freeze up again with “record“ running. The file contains the last 13 days of graphics.
I have isolated what I think is the right file in the temp folder (there are 655 of them). Now that I have the file how do I send it to you?
ID: 38047 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 38049 - Posted: 2 Oct 2009, 17:02:50 UTC
Last modified: 2 Oct 2009, 17:03:59 UTC

Thanks Jim.

I\'ve sent a private message with a suitable e-mail address. The PM should appear by following the \'Inbox\' link at the top-right of the thread pages.

I\'ll process the file and plot the result on a map, or increment the count if the model has crashed in the same place as a previous model.

PS hagar\'s file seems to have arrived ok - thanks.
ID: 38049 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 38050 - Posted: 4 Oct 2009, 18:27:29 UTC
Last modified: 4 Oct 2009, 18:28:19 UTC

Thanks to JIM and hagar for a slab and a mid-holocene iceworld. These have now been processed through the plotting software: one froze in phase 1 and one in phase 4 - but both froze at the same location off the west coast of North America. The image earlier in this thread has been updated accordingly, with points #14 and #15 at the south-eastern extreme attributable to these latest models.

PS The single AMD/Windows model I\'ve received so far - thanks mo.v! - froze in the Mediterranean, so adding to that category might be productive if anyone has such a thing (an AMD/Windows iceworld speeds up when it freezes, so it\'s easily missed).

Keep \'em coming.
ID: 38050 · Report as offensive     Reply Quote
Profile mo.v
Volunteer moderator
Avatar

Send message
Joined: 29 Sep 04
Posts: 2363
Credit: 14,611,758
RAC: 0
Message 38051 - Posted: 5 Oct 2009, 1:30:44 UTC

Iain, I have another. It\'s a HadSM slab running on the Intel, slow-processing in phase 3. I\'m running a couple of beta models on the same machine and they\'ll have to be finished first for Tolu. After that I\'ll restore the backup.

I think you should copy this appeal to the News thread of both forums to get more results particularly on AMD.
Cpdn news
ID: 38051 · Report as offensive     Reply Quote
Profile Iain Inglis

Send message
Joined: 9 Jan 07
Posts: 467
Credit: 14,549,176
RAC: 317
Message 38053 - Posted: 5 Oct 2009, 8:52:52 UTC

Thanks, Mo. Now I\'m back from holiday, I should really write a coherent set of instructions so that no-one runs into any trouble when restoring a backup. (However, my first priority this weekend is to fill in all the holes dug by foxes while we were away - they dug into and emptied the compost bin among other incursions. Grrr.)

I\'m running out my accumulated mid-holocene models and it\'ll be slabs from then on ...
ID: 38053 · Report as offensive     Reply Quote
1 · 2 · 3 · 4 . . . 7 · Next

Message boards : Number crunching : Iceworld Appeal

©2024 climateprediction.net