Yes, all the models help and the researchers are very pleased.

---

Cpdn news

Welcome to the Project and the Boards, Joe.

Credit is awarded for each Model year, 25920 Time Steps, 0.625%. After the report is sent home, a daily batch run sorts it and adds credit to your account.

Be sure to stop the Model and boinc from time to time and make a backup of the entire folder to protect your investment. Les\' How to backup/restore:
http://bbc.cpdn.org/forum_thread.php?id=1952#8299
http://www.climateprediction.net/board/viewtopic.php?p=43422#43422

Still not a \'set and forget\' project and am being sure that I wont be doing \'back up\' for lack of proper recovery routines if something goes wrong with a model. Is CPDN planning to do anything in development with Berkeley to fix this problem? About 6 crashed models so far as backing up and restoring means that other project time shares and work are being sacrificed!

I\'ve asked the Boinc developers at Berkeley several times (trac, bugzilla, and directly) for some way to make backup and restore better, but they\'re not interested (not relevant to the majority of projects).

---

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

Good evening!

May be this work not for BOINC developers. :-)
My BOINC run under WindowsXP and in last two hours I wrote this pair of batch files:
1. backup.bat - \"core\" of command that run backup of BOINC folder;
2. backup_start.bat - batch file which start backup.bat with parameters, which the user can change.

This files start WinRAR archiver, that in my (sample) case create archive with name like BOINC_2008-01-23.RAR in D:\\work folder and start BOINC Manager with --systray parameter.

BACKUP.BAT:
REM =======================================================
START /LOW /WAIT \"BOINC backup\" \"%~3\\WinRAR.exe\" a -r -s -m5 -rr5p -t -agYYYY-MM-DD %2 %1
START \"BOINC\" \"%~1\\boincmgr.exe\" --systray
REM =======================================================

BACKUP_START.BAT:
REM =======================================================
backup.bat \"C:\\Program Files\\BOINC\" \"D:\\work\\BOINC_\" \"C:\\Program Files\\WinRAR\"
REM =======================================================

Parameters:
1. Path to BOINC folder without \"\\\" on the end;
2. Template of backup file name. In this sample - \"D:\\work\\BOINC_\". WinRAR adds current date as YYYY-MM-DD;
3. Path to WinRAR folder.

If user have another path to folder they can simple edit backup_start.bat.
I create a shortcut \"BOINC backup\" to backup_start.bat (placed in BOINC folder) and before run it manually exit from BOINC Manager (do not know how to finish the application from command line :( ).

I like WinRAR (www.rarlab.com) because this have an impressive list of command line switches. If somebody will have questions - I shall try to answer.

---

The important thing about backing up is to make sure that no part of BOINC is running, otherwise the process may fail on backup or fail on restore. A second thing is to synchronise the backups with the checkpoints, which is much harder.

What Mike was asking the BOINC developers to consider was a cross-platform backup integrated into BOINC Manager, which they are unlikely to do so long as the majority of BOINC projects have short work units.

In any event - and this may be heresy! - the important thing is to prevent crashes, not fix them. :-)

The important thing about backing up is to make sure that no part of BOINC is running, otherwise the process may fail on backup or fail on restore. A second thing is to synchronise the backups with the checkpoints, which is much harder.

What Mike was asking the BOINC developers to consider was a cross-platform backup integrated into BOINC Manager, which they are unlikely to do so long as the majority of BOINC projects have short work units.

In any event - and this may be heresy! - the important thing is to prevent crashes, not fix them. :-)

For full set of backup and recover functions for BOINC need an integration with DBMS....

P.S. AND \"BOINC Backup Server\" Software for \"huge cruncher farms\"? ;D

---

Hello, I have a problem: 25 Jul 2008 I casually stops climateprediction process throw windows tasks manager and it has caused \"computation error\". I restored BOINC folder from backup and my tasks calcultaion had continued (however with loss of 2 percent). But in my account all 4 tasks (i have 4 cores) gets Time reported or deadline = \"25 Jul 2008\", Server state = Over, Outcome = Client detached, Client state = New. After several days when the lost percent was calculated again, the credit has again started grants. Whether there is a sense to continue calculation of tasks or they will be rejected by a server in any case? I mean whether there will be a real profit to climateprediction project from this tasks?

P.S. sorry for bad english

1) NEVER stop the program by using Task Manager!
Use the Exit command in BOINC manager.

2) The messages that you got are normal when restoring from a backup.
The data will be OK. It will be accepted by the server, and also by the project people.

Hi Fever

There are some very useful items in the project READMEs about how to keep models going to the end. Get to them via the link in my signature. In the collection about crashes and problems, look at item #6 by Mike Mars who lists what to do & what to avoid.

(Backups to the rescue again! Well done for having one!)

---

Cpdn news

Thanks guys

Hi. I just started running 2 simulations. I\'m curious at about what time during the day the server will update credit. Also, I\'ve sent 3 trickles so far, and they all request 0 seconds of work and report 0 tasks complete. Is this normal?

Thanks!

Hi Eric, welcome to CPDN.

While your model\'s running your computer doesn\'t need any new tasks, so BOINC realises this and requests 0 seconds of new work. BOINC detects how much work the computer can cope with. And BOINC will only report a completed task when your model has finished. So yes, this is all normal.

You can see details of the work you\'ve done by clicking on your forum name. In fact you can click on anyone\'s forum name to see what their computers and models are doing.

---

Cpdn news

Hi all, newcomer to CPDN here with a couple of Questions.

Just got a box running one of the shorter units

workunit

From what I can glean from this thread, there are no trickles until 1 model year is complete. As these units only complete 1 year, does that mean that they will try and run 370 hours without checkpointing/trickling?

As for backing up, is it worth it on this type of wu?


Thanks in advance

Grant

From what I can glean from this thread, there are no trickles until 1 model year is complete. As these units only complete 1 year, does that mean that they will try and run 370 hours without checkpointing/trickling?

Not quite. The HADAM3 models that you are running do run for one year, but submit one trickle every 720 timesteps, adding up to 72 trickles in all.

As for backing up, is it worth it on this type of wu?

Some people back up a lot, some a bit and others not at all. My view is that the amount of backing up depends on how the machine is used. If it\'s used by one person for pretty standard things and shut down carefully, then one or no backups is a sensible strategy. If the kids are thrashing the machine with shoot-em-ups and the power supply is lousy, then crashes are likely - so, if you want to finish a model, then at least one backup is a good idea.

For the longer HADCM3 models, then more backups would be appropriate.

(I take one backup on downloading and no others.)

Just to add a bit:
The hadam models only return zip files at the end of the model.
These are different to the trickles.

Different model types; different checkpoints, trickle times, & zip uploads.

---

Backups: Here

Credit didn\'t run last night. Same thing happened last week. Is this a new periodic operating procedure?

*** Edited - spelling! ***

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

It\'s just one of those things.
You might want to keep an eye on this thread in Number crunching, which is where these problems get discussed.

---

Backups: Here

Okay Les, thanks.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.