Re: [rdiff-backup-users] rdiff-backup unexpected problem.

[dean gaudet]

On Tue, 15 Feb 2005, Brice Figureau wrote:

> Since a few days (our Internet connection was accidently down), I can't
> backup one of the machine (the other seems to be fine), and all I get is
> the following error:
> 
> Read from remote host server3: Connection reset by peer
...
> A tcpdump trace show that the remote server closes the TCP connection (RST), 
> thus the Connection reset by peer.
> Anyone have an idea of how I can fix that ?

i have this problem on and off and it typically clears up in a few days.  
my wild guess is that it's related to heavy NAT traffic (my NAT is just 
another linux box) when there are dozens of bittorrents going... but i've 
never proved this wild guess true or false.  i think the problem is 
*somewhere* in the networking as opposed to in the rdiff-backups though...

i generally just run rdiff-backup again and again until it makes it the 
whole way through the backup.

i've tried a bunch of stuff -- including enabling ssh protocol level 
keep-alive (i.e. place "BatchMode yes" into your .ssh/config entry if 
you're using such a thing).


> How can I tell the remote site to verbose log everything it does ?

i've run tcpdumps and strace's on the remote side and still haven't 
figured it out...


> Running on a virgin directory seems to exhibit the same problem.

i've had success blowing away the rdiff-backup-data subdir, doing an 
rsync, then doing an rdiff-backup --force... the rsync can always be made 
to complete if you run it enough times (i.e. use -avP --delete ... make 
sure you get your path specs right by using -n first).

using volume/fs snapshots can help because you can be sure the image 
you're backing up isn't changing...

it can be frustrating.

-dean