Folding Forum

Right, but not only working directory needs to be different: the "background" instance of Langouste needs to be executed on remote machine. That's where the challenge is...

V7 won't change 2h upload deadline which is what Klue is aiming to mitigate.

The idea is that results are relayed to the middle man in a state-less manner. Think of it as automatic sneakernetting.

I think there's a bit more that would need to be done. Problem with listening on 0.0.0.0 is that when remote client connects to Langouste, Langouste is not able to determine (remote) cllient's PID and, consequently, working directory of the client -- decoupling is turned off and Langouste acts like ...

Right, though you'd still need a method to reliably relay the data between two locations. As both locations are yours 2-hour limit would not apply. Is your link just borderline slow or unreliable (drops) ? Running remote instance of Langouste could be a bit challenging. In your environment.... Would...

Hi bollix47, Actually, 0.7-pre11 hasn't been tested w/V7 and you... have found a bug (which causes DLB engagement code not to start under V7). I've fixed it in 0.7-pre13 -- http://darkswarm.org/thekraken-0.7-pre13.tar.gz, so feel free to give it a shot -- should work much better. Please follow up vi...

It depends on what you are trying to accomplish...

Currently, decoupling requires an instance of Langouste per machine.

Upload capping could be made to work with one instance per... subnet (or whatnot)
but you'd lose decoupling feature.

What is your use case?

31.5GB when I stopped it. total used free shared buffers cached Mem: 33014688 31501092 1513596 0 32 3851632 -/+ buffers/cache: 27649428 5365260 Swap: 2097148 0 2097148 Yes, as you can see, no progress was made in over 40 minutes. [04:53:52] Initial: 0000; - Receiving payload (expected size: 9122) Ho...

Mis-sized P6901 that, not surprisingly, doesn't simulate at all and produces usual memory leak instead. [04:53:51] Connecting to http://assign.stanford.edu:8080/ [04:53:51] Posted data. [04:53:51] Initial: ED82; - Successful: assigned to (130.237.232.237). [04:53:51] + News From Folding@Home: Welcom...

At [H], we've got G34 performance checklist: http://hardforum.com/showthread.php?t=1669698

To answer your question, no, these settings should not affect Langouste's operation in principle.

Working on FTP setup, will PM the details.


Thanks,
tear

Thanks for the report, wombat. Do you still have complete client directory? (w/unsent results, that is) If so, can you please keep it (make a copy or whatnot) and make it somehow available to me? (could set FTP up for you or something) I'd like to understand what happened here but if we wipe everyth...

Gotta remove machinedependent.dat as well -- otherwise unit boomerangs in no time....

Yeah, I don't think that's going to work. Result writeout isn't spread across duration of the simulation.
It's one bulky write at the end of it...

Yes, but only as regular HTTP proxy (plus rate limiting).

You would need to recompile it (with standalone support) and, once in "non-local mode",
you would no longer be able to use its upload/download decoupling feature.

What is the use case you're after?


tear

Excellent. Thanks for keeping finger on the pulse.

Can someone please check? Thank you. [05:38:36] Project: 6903 (Run 10, Clone 17, Gen 34) [05:38:36] - Read packet limit of 540015616... Set to 524286976. [05:38:36] + Attempting to send results [February 11 05:38:36 UTC] [05:38:36] - Reading file work/wuresults_06.dat from core [05:38:36] (Read 2225...