Message boards :
Number crunching :
hadcm3n backup restore -- backups can save weeks or months
Message board moderation
Author | Message |
---|---|
Send message Joined: 31 Aug 04 Posts: 391 Credit: 219,896,461 RAC: 649 |
these hadcm3n models run a long time -- 400 - 700 hours or more Having a backup to restart from can save weeks of rerun or reissue time - and we're working with short deadlines and later models that depend on what we're running now. So if your machine fails, whether disk space, power failure, whatever (not a model problem, but a local machine problem) you can easily restore from a backup and restart -- IF you have a backup. . Please refer to this old posting If your restart is successful, the earlier failure may have been reported by BOINC. You will see on your account page some kind of status that the task has failed. Not a problem, keep on crunching. After restore and restart the 'failed' model will continue to record trickles and upload the all-important data files regardless. The restarted jobs will NOT be wasted, whatever you see on the "tasks" page like "Error while computing" or "Client detached" like here where my machine ran out of disk space because I was doing a big upgrade. Note the trickles picking up again after the restart caught up with its earlier point of failure. So -- please consider doing backups especially when running these long-running and very valuable models. And don't hesitate to post if you have questions or problems with backup/restore Invite comments from moderators or experts if any of this information has changed - thanks. Thanks -- [edit] there may be some changes to what folders you want to back up depending on Windows and BOINC version - for linux its just the BOINC folder. Eric "Die Welt is alles was der Fall ist" -- wrote Ludwig Wittgenstein |
Send message Joined: 31 Dec 07 Posts: 1152 Credit: 22,363,583 RAC: 5,022 |
Welcome to the club. I have posted several times over the years about the benefits of making frequent backups of these long models. As you say they can save you weeks of crunching. I make one almost every day. It’s fast and simple. Once you get the hang of it, it only takes a few minutes. To backup a model in the 6.xx.xx version of the boinc manager: 1. Make a folder in “My Documents” and add several sub-folders inside to receive your backup. 2. Suspend WU and exit the Boinc manager. 3. Click “Computer” in the Windows Start Menu. 4. Double click on drive “local dick ( C )”. 5. Navigate to the ProgramData and open it. (Note: Windows hides this folder by default, so it will be necessary to make it visible the first time you do this. See below.) 6. Locate the “Boinc” folder and open it. 7. Copy the entire contents of the folder to one of the sub-folders you made in “My Documents. ” 8. Close the Boinc folder and the ProgramData folders and restart boinc manager. Your done. To restore a model in the 6.xx.xx version of boinc manager. 1. Exit boinc manager. 2. Open ProgramData and Boinc folders. 3. Delete entire contents of Boinc folder. 4. Open most recent backup folder. 5. Copy entire contents of this folder to “Boinc” folder in ProgramData folder. 6. Restart computer to clear any remaining problems. To make the ProgramData folder visible type “folder options” in the search box and click on it when it appears in the menu. Then click the “view” tab. Find “hidden files and folders” and click “show hidden files, folders and drives.” Click “apply” and then “OK”. ProgramData will now be visible. As Eirik stated this can save you week or months of crunching (and a great deal of frustration) and get you back on track to a successful completion. |
Send message Joined: 17 Nov 07 Posts: 142 Credit: 4,271,370 RAC: 0 |
Just to emphasize one point that Jim made: a backup made while any Boinc work is still running will most likely NOT be usable. It's important to suspend all Boinc work before making the backup. This applies mainly to scheduled (automatic) backups, for example to an external hard disk. It's best to exclude the Boinc folder proper from scheduled backups, and include just the backup folder that you made when following Jim's instructions. |
Send message Joined: 3 Oct 06 Posts: 43 Credit: 8,017,057 RAC: 0 |
I think it might be worth mentioning that this method of restoring a BOINC backup will also trash the other BOINC tasks, if you're running more than one BOINC project. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
My sticky about backups is here. The BOINC core client must be EXITED from before making a backup, so that none of the model's many files remain locked. There are some indications that suspending the BOINC core client near the end of a model year will cause the hadcm3n models to crash. So best to avoid doing this from, say, mid November to mid January. Backups: Here |
©2024 cpdn.org