How to backup your F@H files for recovery

[ThunderRd]

This will be a tutorial on what I have found out regarding F@H backups, and their value. If you are running the SMP core you may have already seen that *sometimes* a WU just fails. If you are overclocking you may take a look at your temps, etc to see if you can find the culprit. The problem is that the WU is gone by this time and after you make your adjustments you may well get a different unit. All of them behave differently, so it's not a reliable check.

What if you could roll back your WU to a point prior to where it died, make your adjustments in BIOS, or whatever, and start it up again? Then you could see how the SAME WU behaves with amended settings.

I am not overclocking any of my business machines, and I have seen a number of WUs die for no apparent reason. Once there was a spurious IP conflict on one machine, and 12 other boxen hung on the network. I found the problem 5 or 6 hours later, and lost some of the WUs as well as 60-70 hours of production time. That's an example. It was that event that inspired me to think about a backup strategy that would protect the work I had already done, with minimum loss of time. Remember that I have nearly 40 machines on the SMP core. I needed something, because the client isn't exactly what we'd call 100% stable at this time.

For those of you already familiar with Windows Backup this may be elementary, but for others it may save you some fiddling. WBU is not exactly the most intuitive of programs, so here goes:

Open WBU to the "Schedule Jobs" tab, and click Add Job. This brings up the wizard. Next through it. Dialog window appears, click "back up selected." Find your F@H directory in the next window. Personally, I backup the entire directory (and *work* sub-directory), although there are some files that aren't necessary. Tick the directory in the left-hand pane. Now, this is important - after ticking the checkbox you must also CLICK ON THE DIRECTORY TITLE; when it appears in the right-hand pane you are good to go ahead.

Next, in the right-hand pane, UNTICK *smpd* and *msiexec*. This isn't really important, but it will save you a reboot after restoring, because at the very least, smpd will be a running system process that WBU can't restore until a reboot. (Perhaps msiexec as well, if you didn't kill the process.)

Clicking Next will bring up the destination window. Choose the place you want to BU to; I use a BU directory on the drive. Name the file and click save, next.

The Type dialog follows. I use Normal. We're going to delete the BU file after the WU is done, so we're not concerned with disk space. The next dialog is an options window. Tick "Verify Data". Don't tick the other two boxes, especially "Disable Shadow Copy"; this is enabled by default so Windows can back up a file currently in use. Many of our F@H files will be in this state so it has to be enabled. The next window has more options. MAKE SURE you tick APPEND BACKUP. If you don't, we will overwrite the prior backup and you will lose rollback ability, which is what we're after.

The next window is a scheduler, called "When to back up". Click Later, and name the job file. Click Set Schedule. On the Schedule tab, in the Schedule task drop down, click DAILY. This isn't intuitive at all, but bear with me for a moment. Fill in a Start Time about 1 hour in the future. In Schedule Task Daily, make sure it reads 1 Day.

In the same window, click Advanced. Enter Today's date, do not click End Date. Click Repeat Task. Enter "Every 60 minutes" or as frequently as you want. Click the Duration radio button. Enter 24 hours, and OK out of the window. The next window will recap your settings; I had no need to make any other settings, although there are some power management options on the Settings tab. OK out of this window.

A dialog will come up asking for a password. If you are running the SMP client, you must have a Windows logon password set. The SMP client doesn't like blank passwords, and will most likely not work without one. Well, WBU won't work without a password either, and you will have to enter it twice right here. Enter your logon password now. You will get the same "When to back up" window that you saw before. Do nothing except Next to leave the window.

Windows will then ask you for your password again. Enter it and OK out. Finally, you will be presented with a recap window, called Completing the Backup Wizard. Review the details for correctness, and Finish.

Whew.

Now, each hour, WBU will backup your active F@H directory, and it will append each hour's BU to the prior ones. If a WU hangs or breaks, you can use WBU to roll back the directory to recover lost work. Just use the Restore feature. It is a lot more intuitive than the scheduler. Just make sure that it is set to Replace Files, as it isn't by default. Then, in the left hand pane, scroll down to the last BU and count back hour by hour until you know that you are at a point before the stall. Then Restore. If it doesn't work then count back 1 or 2 hours more and try again.

For housekeeping, simply delete the backup file after uploading the finished WU. This way you won't slowly eat up your disk space. It will also be easier to find the correct hour to restore if you need to. You don't want to have more than one WU in the backup file.

For people who have lost WUs for no apparent reason, I can say that I have, so far, been successful in completing about 90% of stalled WUs after rolling back.

Hope this helps some of you.

[robbie]

Thank you for the tip!
We may wanna sticky this one.

[Strongwolf]

Excellent!!! That would save me a headache and some bad words. Thanks.

I hope you won't mind I copy the tutorial and paste it in my documents.

[Samuknow]

Front Page

http://www.aoaforums.com/frontpage/content/view/3189/1/


Thanks Robert. This is good stuff.

[ThunderRd]

Bump it so it doesn't drop off the first page. Lotsa folks with problems right now.

This routine has helped me to recover many wu's since I began using SMP.

[Samuknow]

I did put this up on the F@H front page.

I will check to see if it is getting any hits.