Backup Central

dean gaudet <dean-list-rdiff-backup < at > arctic.org>
wrote the following on Sun, 4 Apr 2004 17:11:05 -0700 (PDT)

i had rdiff-backup 0.11.0 running smoothly for over a year and
recently started having problems with it failing to complete a
backup. figuring i was long overdue for an upgrade i moved to deb
unstable recent which is 0.13.3.jgoerzen-3 ... i'm still having
problems with the backup failing to complete.
...
self.conn.os.link(linkpath, self.path)
OSError: [Errno 2] No such file or directory
...
rdiff-backup already knows that link is going to fail, and yet it
goes ahead and does it -- but more importantly is that it results in
the traceback above. the final write is the beginning of it bailing
out and printing the traceback.

it would be preferable not to bail out on this error -- i'm pretty
sure what's happenned here is that the .../155/61018/ directory has
disappeared on the source (yeah i'm bad and backing up from a
non-snapshot source).

I think I just fixed this in CVS. But I'm still not sure how you
managed to run into this problem repeatedly. To replicate it I had to
make three hardlinked files, and then delete the first after it had
been stat()ed and before it starts getting read.


--
Ben Escoto

dean gaudet <dean-list-rdiff-backup < at > arctic.org>
wrote the following on Sun, 4 Apr 2004 17:11:05 -0700 (PDT)

i had rdiff-backup 0.11.0 running smoothly for over a year and
recently started having problems with it failing to complete a
backup. figuring i was long overdue for an upgrade i moved to deb
unstable recent which is 0.13.3.jgoerzen-3 ... i'm still having
problems with the backup failing to complete.
...
self.conn.os.link(linkpath, self.path)
OSError: [Errno 2] No such file or directory
...
rdiff-backup already knows that link is going to fail, and yet it
goes ahead and does it -- but more importantly is that it results in
the traceback above. the final write is the beginning of it bailing
out and printing the traceback.

it would be preferable not to bail out on this error -- i'm pretty
sure what's happenned here is that the .../155/61018/ directory has
disappeared on the source (yeah i'm bad and backing up from a
non-snapshot source).

I think I just fixed this in CVS. But I'm still not sure how you
managed to run into this problem repeatedly. To replicate it I had to
make three hardlinked files, and then delete the first after it had
been stat()ed and before it starts getting read.

in the end i'm convinced that this link error was temporary (although i
still don't know what i did to cause it ... it seems the main problem
i've been having with rdiff-backup is not rdiff-backup's fault -- it's
because of using bittorrent -- and bittorrent appears to act like a DoS
on my NAT causing it to overflow its connection table... and somehow
the rdiff-backup ssh session through the NAT gets dropped frequently
when this happens. at least that's my current working hypothesis :)

but thanks for looking into the link error.

oh hmm you know -- i use qmail, which does a lot of hardlinking in
/var/qmail/queue and behaves somewhat like you describe.

-dean