Re: [Duplicity-talk] gpg io error

[Mitchell Perilstein]

Hi Folks,

>>  File "/usr/lib/python2.3/site-packages/duplicity/gpg.py", line 115, in
>>  close
>>    self.gpg_process.wait()
>>  File "/usr/lib/site-python/GnuPGInterface.py", line 639, in wait
>>    raise IOError, "GnuPG exited non-zero, with code %d" % (e << 8)
>>IOError: GnuPG exited non-zero, with code 131072

Just wanted to give an update on this issue.  The bad news for the list is I wasn't able 
to narrow it down to help solve the problem.  The good news for me is by various 
workarounds and hardenings I was able to get backups running again and have not seen the 
error recur:

 1. Clean up the source area.  There may have been some old or inappropriate symlinks 
hanging around and there was some junk that didn't need backing up.  I'm pretty sure this 
solved the I/O error ultimately, but just in case, read on...

 2. Partition the source area into smaller chunks to be performed on different nights 
(cron + scripty bits).

 3. Manually perform a full backup to a local file:/// target, then copy that to the 
remote backup host for each chunk.

 4. Restart the cron+scripty backups going to the remote host scp:// target, which holds 
the newly uploaded full chunks.

 5. A major improvement we've seen at some of our sites is a wrapping layer around ssh 
and scp which performs a fallback and retry in the case of a slow or unreliable internet 
connection acting up during backups.  Instead of failing the whole backup, my ssh wrapper 
will keep retrying, timing out, and waiting longer times until it succeeds.  From 
duplicity's perspective, it just runs scp once and waits until eventually it happens or 
fails after a lot of trying.

The SSH wrappers are an effective hack, and we plan to share them with the list shortly, 
but ultimately, this points out a few duplicity improvement wishes for a new thread.

Best regards,
--
Mitchell Perilstein
Partner
ACE Technology Group, LLC
http://www.acetechgroup.com
(866) 229-1543 x11

mitch.vcf
Description: Vcard