climateprediction.net home page
Strange behaviour in statistics

Strange behaviour in statistics

Questions and Answers : Unix/Linux : Strange behaviour in statistics
Message board moderation

To post messages, you must log in.

AuthorMessage
old_user431158

Send message
Joined: 12 Feb 07
Posts: 9
Credit: 51,485
RAC: 0
Message 31593 - Posted: 4 Dec 2007, 20:17:09 UTC

Usually the statistics graph on my model shows a steady increase with trickles about every three days. Over the last few days this has changed with two successive trickles showing no increase on two occasions:-

27 Nov 29300
28 Nov. 29550
30 Nov. 29550
1 Dec. 29800
4 Dec. 29800

These numbers seem to agree roughly with the trickle records on the Web page.

Is this normal?

Many thanks.


ID: 31593 · Report as offensive     Reply Quote
Profile MikeMarsUK
Volunteer moderator
Avatar

Send message
Joined: 13 Jan 06
Posts: 1498
Credit: 15,613,038
RAC: 0
Message 31594 - Posted: 4 Dec 2007, 21:41:11 UTC


From your result\'s page (http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=6401667), the recorded trickles are:

30th Nov
27th Nov
24th Nov
22nd Nov
13th Nov

Note that trickles will only be recorded if it is the first time that the model has sent them. It is possible for the model to rewind in certain circumstances, which mean that duplicate trickles can be sent. However, the s/ts times look stable, which suggest that there haven\'t been any rewinds.
I'm a volunteer and my views are my own.
News and Announcements and FAQ
ID: 31594 · Report as offensive     Reply Quote
old_user431158

Send message
Joined: 12 Feb 07
Posts: 9
Credit: 51,485
RAC: 0
Message 31595 - Posted: 4 Dec 2007, 22:14:15 UTC

Thanks for that Mike.

I wonder if it is my fault.

I shut the process down after I\'d seen the \"Scheduler RPC Succeeded Message\" in the messages window but wonder if this appears before the files have been uploaded. In other words the \"succeeded\" just means that the request to upload has been granted, not that the upload has been successful. Certainly in stdout.txt it looks as if the zip file is built after the RPC succeeded message is logged.

Do you know if this is the case? If it is then the program would have needed another shot at sending the trickle message when I restarted it.


ID: 31595 · Report as offensive     Reply Quote
Profile MikeMarsUK
Volunteer moderator
Avatar

Send message
Joined: 13 Jan 06
Posts: 1498
Credit: 15,613,038
RAC: 0
Message 31596 - Posted: 4 Dec 2007, 22:19:15 UTC
Last modified: 4 Dec 2007, 22:21:08 UTC

If you shut it down immediately after a trickle, it often resends the trickle again - this is harmless.

(Trickles are sent on the model date 4th December, and the previous checkpoint is at 30th Nov. So days 1st Dec to 4th Dec are recalculated, and the trickle sent again).

The best time to stop the model is immediately after a checkpoint (usually every 10=20 minutes or so). I\'ve set up my \'cc_config.xml\' file so that a checkpoint message appears on the messages tab, but you can also see the checkpoint on the \'show graphics\' screen (\'Z\' to remove the grey bar, and then \'8\' to get the checkpoint countdown).
I'm a volunteer and my views are my own.
News and Announcements and FAQ
ID: 31596 · Report as offensive     Reply Quote
Les Bayliss
Volunteer moderator

Send message
Joined: 5 Sep 04
Posts: 7629
Credit: 24,240,330
RAC: 0
Message 31597 - Posted: 4 Dec 2007, 23:06:41 UTC


A simple way to check is to look at the BOINC\\projects\\climatepredition folder.
THEN enable Network access, and watch both the Messages tab and the folder display.

Usually the file(s) disappear and the message says Succeded at about the same time, but sometimes one takes a few seconds longer than the other. And which happens first is also random.

And sometimes there is a problem with the transfer, and, although the Succeded message is displayed, the files remain in the folder until the next trickle is sent. (I think that BOINC resets it\'s \"sent\" flag when it gets the Success message, so it doesn\'t know that the trickles are still there.)

What you\'ll see in this case is that the icon image used for the trickles changes, and then they stay like that instead of disappearing. I\'ve had it happen a few times recently, but it\'s not a problem in the long run.
Unless one is desparately waiting for the trickles to get there so that one gets credit for them before the stats are exported. :)

ID: 31597 · Report as offensive     Reply Quote

Questions and Answers : Unix/Linux : Strange behaviour in statistics

©2024 climateprediction.net