Linux - Par2 repair fails 0.4.6 [SOLVED]

[mollydog]

Ever since installing 0.4.6 there has been problems with lots of _FAILED_ downloads. I tried to find out why today.

Looking in the logs it seems par2 verify always fails but files unpack OK when repair is not needed, obviously failing when it is.

eg.
One which unpacked OK..
2009-02-10 09:01:41,184::INFO::[newsunpack] Par verify failed on *removed*.vol00+01.par2!
2009-02-10 09:01:41,309::INFO::[newsunpack] Unrar starting on  *removed*
2009-02-10 09:01:41,310::INFO::[newsunpack] Extracting rarfile  *removed*.rar (belonging to  *removed*) to  *removed*_UNPACK_ *removed*
2009-02-10 09:01:53,579::INFO::[newsunpack] Unpacked 1 files/folders in 12.3s

One which failed..
2009-02-10 09:30:09,817::INFO::[newsunpack] Par verify failed on  *removed*.vol00+01.par2!
2009-02-10 09:30:09,848::INFO::[newsunpack] Unrar starting on  *removed*
2009-02-10 09:30:09,849::INFO::[newsunpack] Extracting rarfile  *removed*.rar (belonging to *removed*) to  *removed*_UNPACK_ *removed*
2009-02-10 09:30:11,225::WARNING::[newsunpack] ERROR: CRC failed in  *removed*
2009-02-10 09:30:11,313::INFO::[newsunpack] Unrar finished on  *removed*


par2 is found in /usr/bin (par2cmdline version 0.4) and the files repair correctly when used manually with par2repair.
::INFO::par2 binary... found (/usr/bin/par2)

[switch]

Do the downloads contain any weird characters or accents?

I would try and download the first job again with no postprocessing set (set the mode to blank in the dropdown). Then manually run par2

Code: Select all

par2 r path/to/file.par2

And then if you can save the output, or pipe it to a text file and send it to [email protected] and we'll see if it is a problem with sabnzbd.

[mollydog]

It seems to fail verification of every single download, no particular one. Downloads which don't require repair pass through and unpack OK while the damaged ones fail CRC check.

Running 'par2 v whatever.par2' on _FAILED_ downloads fails verification but 'par2 r whatever.par2' does repair them without problems. I thought it may be a problem with white space in folder names but using the switch to replace spaces with underscores doesn't improve matters.

I have it set to carry out repair and delete. Nothing wierd about the names.

[switch]

I've run your failed log file through some test scripts I use and it comes out ok. Are you positive this has just started happening with 0.4.6? What happens if you go back to 0.4.5?

[mollydog]

Ummm nope, not 100%

I'll go back to 0.4.5 etc  and see what happens. The failures did seem to start all of a sudden though.

[mollydog]

Well Ive gone as far back as 0.4.3 and as far as I can see it seems the same on all of them 

It must be something to do with the system causing it. Python is v2.5.2.

Any ideas?

[sander]

What if you start clean:
1) (re)move ~/.sabnzbd/sabnzbd.ini
2) download and unpack a fresh SABnzbd 0.4.6
3) startup the fresh SABnzbd, fill out one newsserver, download a fresh and see what happens

If the par2 still fails, I could check the NZB on my own Ubuntu system to see what happens.

[shypike]

I have just downloaded your "betamax" job with 0.4.6.
I run Ubuntu server 8.04 on an old PentiumIII pc.
I downloaded it once with the standard repo par2 (the classic single-core version)
and once with chuchusoft's multi-core version.

Both downloads didn't have any repair problem.

As I do not have your NZB file, I had to get it from binsearch.info.
Please email your NZB file so that I can check again.

Somehow the par2 you are using is not compatible with SABnzbd.
Strange, because there are only 2 popular variants.

Can you give some more info about your system?

[mollydog]

Thanks for your help. Problem solved! What you said about the multicore version made me think. I had -t0 set as extra par2 parameters which once removed seems to have cured it! Silly error on my part by the looks of it. Could this kind of thing be made to show up in the logs, reporting any error messages?

Thanks anyway, you've been very helpful.

[shypike]

Should have thought of that too.
The old par2 (single-core) program doesn't understand the -t0 parameter
and just panics.

We should actually analyze par2 program's reaction better, instead of just
assuming that any error exit means "cannot repair".

BTW: you need to be careful with nice and ionice parameters too.
Incorrect parameters there will have the same result.

[mollydog]

Thanks, I'll look into getting the multi threaded par2 installed 

I hadn't tinkered with the others so will probably leave well alone and not fix what isnt broken (anymore)!