Page 1 of 1

Almost always failing...

Posted: September 13th, 2011, 4:54 pm
by nkr
Hello all,
I'm having an increasingly frustrating issue with the otherwise excellent sabnzbd. It's set up as a service on WHS2011, running 0.6.9, with the files stored on a Drive Bender pool but temporary folders on a dedicated physical drive. Almost all downloads fail, coming up requesting more par2 blocks, etc (not just 1-2, but usually 20+, even 80+ for a file that's around 10GB). The downloads are initiated by Sick Beard. The funny thing is it all works perfectly fine when I use Unzbin on my own PC (W7 x64) with the same NZB's which I copy from the WHS2011 machine.

I have pretty much run out of ideas as to what I can try. I can happily clear the logs and do a test, and supply the log file? Anything else I can try / change / etc?

Thank you very much for any help!

Re: Almost always failing...

Posted: September 14th, 2011, 2:58 am
by shypike
Please email some NZB files to bugs at sabnzbd.org
Also how old they were at the time of download on your server?
(If you try them later on your PC, the Usenet server might have caught up by then).
Did do ever do a "Retry"?
Which Usenet provider do you use?

Re: Almost always failing...

Posted: September 15th, 2011, 10:14 am
by nkr
Done, thank you!

Re: Almost always failing...

Posted: September 19th, 2011, 3:29 pm
by shypike
I already mailed you too.
Without the NZB itself, there's not much I can do.

Re: Almost always failing...

Posted: September 21st, 2011, 1:19 pm
by shypike
Judging from the log, all files are read.
It also looks like everything is complete. Not 100% sure because debug logging is off.
So far the only thing I can think of is that par2 screws up.
Did you try the NZB again on your server?
How old was the post when you first tried to download it on your server?

Re: Almost always failing...

Posted: September 27th, 2011, 9:37 am
by kmdm
I just want to suggest a memory check (memtest86+, prime95 blend test)...

I recently had a very similar problem (and symptoms) and it turned out the memory I'd swapped in whilst re-building was faulty and causing no end of problems trying to verify, par2 repair and/or decompress...

OT: I ran mprime (Ubuntu server, not windows) with the blend test and it error'd out pretty quickly, switching to the small FFTs test and it passed just fine. A classic sign of faulty memory which held up to be the case.