Continue bigadv WU in another distro?

[theGryphon]

Hey guys,
Would it be possible to transfer the WU files to another distro and continue working on it there? I mean, would it successfully upload without any problems with the servers?

 

[ChelseaOilman]

You need to transfer more than just WU files. You need to also transfer the queue.dat file and machinedependant.dat at a minimum. Then don't be surprised if you only get base points after you submit the WU. There are lots of posts in the FF asking why did I only get base points. Then after investigation the person admits they sneakernetted the WU. PG has so many checks in place to prevent people from cheating that it's hard not to get tripped up.

 

[Grandpa_01]

It used to be possible to transfer the entire file from 1 rig to the other run the file / folder until it was 99% then transfer it back to the original machine let it finish and send it. But I do not know if that is possible any more or not. VJ has said they get allot of errors when a WU is run on 1 machine then transferred to another and allot of times the results are unusable. So there really is no reason to do it.

 

[theGryphon]

It used to be possible to transfer the entire file from 1 rig to the other run the file / folder until it was 99% then transfer it back to the original machine let it finish and send it. But I do not know if that is possible any more or not. VJ has said they get allot of errors when a WU is run on 1 machine then transferred to another and allot of times the results are unusable. So there really is no reason to do it.

I didn't know about the potential corruption of WUs like that. I thought about an experiment, but I'll leave it alone.

Thanks for the input guys!

 

[tear]

Transferring complete client directory is safe -- no need to complicate things.

If you don't change any files and if your checkpoint is correct, you'll be fine (as long as you don't try
to run the same copy of the client on two different machines at the same time).

 

[Grandpa_01]

Transferring complete client directory is safe -- no need to complicate things.

If you don't change any files and if your checkpoint is correct, you'll be fine (as long as you don't try
to run the same copy of the client on two different machines at the same time).

I am not sure about that tear when talking to VJ one time about it he said that if the machines are not matches it quite often causes corruption to the WU. He said it changes the trajectory of the WU in many cases. That was the reason for Stanford trying to stop sneakerneting. Not because of cheating but because of trajectory corruption.

 

[theGryphon]

Just to clarify, I was thinking about moving the WU (or, the whole client directory) to be folded on the same machine, but a different Linux distro (different kernel, etc.)...

 

[ChelseaOilman]

The best thing to do is use the -oneunit flag and finish your WU, then after uploading install the new distro.

 

[Punchy]

I am not sure about that tear when talking to VJ one time about it he said that if the machines are not matches it quite often causes corruption to the WU. He said it changes the trajectory of the WU in many cases. That was the reason for Stanford trying to stop sneakerneting. Not because of cheating but because of trajectory corruption.

Matches in what way?

That sounds similar, but not the same, as the warning you get if you change the number of threads mid-WU - something about the simulation results not being binary identical - but that message also says the results will still be valid. If they didn't want that to happen, that non-matching-thread detection should simply terminate the WU.

There is implicit randomness in the simulation anyway, so I would find it hard to believe that running the same WU on two different systems using the same number of threads would cause corruption or change the trajectory in a significant way.

Anyway, theGryphon's case as he just specified would work fine.

 

[theGryphon]

Thanks Punchy! Yeah, my case is a bit milder than moving to a completely different system, I guess. Anyway, I didn't do it. It already had the oneunit flag and I just let it finish.