Page 1 of 1

FAH Backup yep or nope

Posted: Mon Apr 02, 2012 2:43 pm
by *hondo*
Anyone know if there's an auto method of backing up the Folding@home-gpu Folder at say 98% complete?

Re: FAH Backup yep or nope

Posted: Mon Apr 02, 2012 4:22 pm
by 7im
Back up your fah folder every 10 minutes. ;)

Other than than, not really. That's what people have done in the past.

Re: FAH Backup yep or nope

Posted: Mon Apr 02, 2012 4:54 pm
by *hondo*
Thanks 7im, TBH I thought I'd sussed out the prior problem of IO errors, but it appears that I occasionally haven't unexpectedly I blew out another one this morning Ahgggggg.

Would it not be possible to enclosed (Something) within FAH that could set off an auto backup if pre planned.

Knowing me, I'd not notice a FAH WU was completed until 30 or more minutes after any prior one's had exploded.. :oops: Therefore I'd restore the new WU just somewhat earlier, which wasn't what I'd got in mind lol.

Thanx for the info though. :)

Re: FAH Backup yep or nope

Posted: Mon Apr 02, 2012 6:05 pm
by 7im
I think it would be more productive to have fah start using multiple checkpoints as the backup. Gromacs supports having more than one checkpoint saved. Just haven't seen that rolled in the the fahcores yet.

Re: FAH Backup yep or nope

Posted: Mon Apr 02, 2012 8:16 pm
by *hondo*
Hi again 7im, I'm really sorry but I wasn't born in an age when PC tuition was taught at school, & TBH I'm completely clueless to understand what it is that you're saying above, can that be done ATM, if so please explain, if not, then thats more than fine by me.

I might be being rather clumsy here with words, but I'm not attempting to be a clever smart ar$e, or in anyway rude.

Re: FAH Backup yep or nope

Posted: Mon Apr 02, 2012 9:28 pm
by 7im
Right now, the fah client only saves one checkpoint file (a place to start if you quit running fah). If that file gets corrupted, the WU has to start over at 0. I assume this is why you want to save a backup at 98%, in case the WU gets corrupted?

If not, please explain why you need a backup at 98%...

To explain more, fah runs using software made by gromacs.org. Gromacs.org has newer software that can save more than one checkpoint file. Almost like having a backup from 5 minutes ago, and another backup from 20 minutes ago. If the 5 minutes one is corrupt, it can go back to the 20 minutes file and try that one. However, Folding@home isn't using this new feature in the fahcores yet.

Re: FAH Backup yep or nope

Posted: Mon Apr 02, 2012 9:48 pm
by *hondo*
7im wrote:Right now, the fah client only saves one checkpoint file (a place to start if you quit running fah). If that file gets corrupted, the WU has to start over at 0. I assume this is why you want to save a backup at 98%, in case the WU gets corrupted?

If not, please explain why you need a backup at 98%...
Yep thats bang on the nail 7im or an hole in one :D

7im wrote:To explain more, fah runs using software made by gromacs.org. Gromacs.org has newer software that can save more than one checkpoint file. Almost like having a backup from 5 minutes ago, and another backup from 20 minutes ago. If the 5 minutes one is corrupt, it can go back to the 20 minutes file and try that one. However, Folding@home isn't using this new feature in the fahcores yet.
OK I hear what you're saying there, and fully understand :) So some day, sooner or later multi backups "just maybe coming soon"

Thanks 7im, very much indeed. :) BTW I can't figure this out, but maybe of help to others, if I stop the CPU folding when the GPU is at 99% these IO errors never happen at all. :? Anyway bed time for me now, so Good Night All :)

Re: FAH Backup yep or nope

Posted: Tue Apr 03, 2012 6:11 am
by bruce
FAH was originally designed to run completely unattended. As 7im said, the checkpoint system is designed to give FAH a chance to resume work after shutting down your system. It is, in fact, a backup, but it doesn't cover every possible situation. If your utility company has a power failure, whether it's at 5% or 95% or you shut down your computer, you can resume work with minimal losses.

Now suppose you have an unstable overclock or somehow there's some hardware problem at 99% -- and assume you've created an external backup system that operated at 98%. Your computer will probably still be running so FAH will report the error. In some cases, you'll be awarded 99% of the credit and in other cases, the results will be corrupted and will be worthless. In both cases, FAH will proceed to download a new WU and will go back to work on a new WU.

Later, you'll notice the problem and you'll want to restore your backup.

Question 1: What do you plan to do with the newly assigned WU? (Please don't say "dump it" because that's bad for science.)

Question 2: What do you expect the server is going to do after receiving an upload that either gives you partial credit or one that simply reports that there was a processing error?

You won't like the answer to question 2. Assuming you restore the backup and process the last 2% of the WU, the server will refuse to accept a second upload of the same WU. Uploading a WU that you created from a backup copy fill generally give you zero points. There are several reasons that the EULA says you are not allowed to manipulate FAH's files.

Every WU is accounted for. It is assigned to YOUR machine ... and restoring a backup disrupts that orderly accounting system, even if you't like to turn back the clock to some earlier time by restoring a backup. It's a bad idea.

Re: FAH Backup yep or nope

Posted: Tue Apr 03, 2012 4:20 pm
by *hondo*
bruce wrote:FAH was originally designed to run completely unattended. As 7im said, the checkpoint system is designed to give FAH a chance to resume work after shutting down your system. It is, in fact, a backup, but it doesn't cover every possible situation. If your utility company has a power failure, whether it's at 5% or 95% or you shut down your computer, you can resume work with minimal losses.
Yep I realise the above completely & utilise this facility when I have the need to re-boot the PC

bruce wrote:Now suppose you have an unstable overclock or somehow there's some hardware problem at 99% -- and assume you've created an external backup system that operated at 98%. Your computer will probably still be running so FAH will report the error. In some cases, you'll be awarded 99% of the credit and in other cases, the results will be corrupted and will be worthless. In both cases, FAH will proceed to download a new WU and will go back to work on a new WU.

Bruce I realise you can’t be expected to remember every word from everyone on here, but how many times do I need to tell you Guys, I’m running now and have been for a fair amount of time at stock values. In no cases have I been awarded anything when these IO errors occurred, just to recap I’ve fitted an higher powered PSU changed the Video card for another brand new one, also I’ve changed to a better ventilated case, and currently just for the sake of F@H I’ve just ordered yet another even better case, a Fractal Designs Define R3 all at considerable expense, I’ve returned & had the PC tested by the supplier (Fortunately FOC) with no faults found.

bruce wrote:Later, you'll notice the problem and you'll want to restore your backup.

Question 1: What do you plan to do with the newly assigned WU? (Please don't say "dump it" because that's bad for science.) .

No, I’d backup that up too & restore it on completion of the prior WU, I don’t just dump Wus, not even if I consider the points accredited to be ridiculously low, as I do with one I started @ around 10.00pm last night & completed @ 9.00am today P7640 which caused my PPD to drop by approx 27% If the only reason I was prepared to folding@Home was PPD, I would have most certainly have dumped that!!
bruce wrote: Question 2: What do you expect the server is going to do after receiving an upload that either gives you partial credit or one that simply reports that there was a processing error?
This is the whole point, the WU isn’t as far as I can see being uploaded @ all to the server from what I can see, there is no activity from the router when these IO errors occur, other than a download the time is too short, so I guess but don’t know, the server is totally oblivious to the fact that the WU has been completed, I might be wrong about that, because I just don’t know.
bruce wrote:You won't like the answer to question 2. Assuming you restore the backup and process the last 2% of the WU, the server will refuse to accept a second upload of the same WU. Uploading a WU that you created from a backup copy fill generally give you zero points. There are several reasons that the EULA says you are not allowed to manipulate FAH's files.
Finally again maybe I’m wrong again here; I can’t see how I’m infringing the EULA at all, I’m not meddling with any internal files in anyway whatsoever, I don’t have either the knowledge or any inclination to do that whatsoever.

If for example I’d completed an Acronis drive image of my Hdd, which included a semi-completed WU, on restoration of the image & completing the semi-finished WU, am I infringing the EULA then, or am I saving a WU from going to waste???

bruce wrote: Every WU is accounted for. It is assigned to YOUR machine ... and restoring a backup disrupts that orderly accounting system, even if you't like to turn back the clock to some earlier time by restoring a backup. It's a bad idea.


How can that be so? If the WU is completed & returned to the FAH server?

Once again I'll make you aware of the fact that if I stop the CPU from Folding these IO errors from the GPU have never occurred, not even once, however I can’t baby sit the PC all day to stop the CPU F@H just because the GPU maybe about to blow out 4 hrs of F@H time, this is a waste to both FAH & a waste of 4 hrs of F@H time to me, let alone the PPD which like it or not, is a major motivating factor for many, me included, fellow Folding@home members to participate in the F@H project.

Re: FAH Backup yep or nope

Posted: Tue Apr 03, 2012 6:08 pm
by bruce
I'm sorry if you were offended by my offering too-much-information.

I may keep track of nearly everything that's said on the forum, but keeping track of who said it and when is beyond my capability so I don't even try ... and I don't choose to search all of the posts you've made in the past and refresh my memory before replying. I'm not singling you out -- I pretty much do that for everybody except maybe those who violate the forum rules.

Also, even though I reply to an individual that asks a question or makes a comment, the forum is a source of information for many, many others -- and I try not to assume that everybody has the same experience and background as you do. I try to help them, too.