[Rdiff-backup-bugs] [bug #23915] rdiff-backup 1.1.5-4 violent crash when backup normal ext3 partition over ssh

[Ludolpif]

URL:
  <http://savannah.nongnu.org/bugs/?23915>

                 Summary: rdiff-backup 1.1.5-4 violent crash when backup
normal ext3 partition over ssh
                 Project: rdiff-backup
            Submitted by: ludolpif
            Submitted on: mercredi 23.07.2008 à 08:40
                Category: None
                Severity: 3 - Normal
              Item Group: None
                  Status: None
                 Privacy: Public
             Assigned to: None
             Open/Closed: Open
         Discussion Lock: Any

    _______________________________________________________

Details:

Situation:
2 OVH servers on Debian Etch, with rdiff-backup 1.1.5-4 installed from
official deb repo.

A cron task launch a bash script at 5:00 am which start this sequence :
rdiff-backup --remove-older-than ${nb_increments}B ${destinationPath}
rdiff-backup --exclude-other-filesystems --exclude "${DEST_BASE}/*"
--preserve-numerical-ids --print-statistics --remote-schema "ssh -p
${destSSHPort} -i ${KEYS} %s rdiff-backup --server --restrict-read-only
${sourcePath}" address@hidden::${sourcePath}
${destinationPath}
with variable replacement, of course.
Backup script duplicate rdiff stdout and stderr in a log file by piping to
"tee -a".

Each server backup the other. There is multiple partitions to backup and I
launch an rdiff per partition, with different cron entry. In one way, it works
perfectly (no error for 3 months, daily backup) for every backup.

In the other way, there is on ONE backup that crashes quasi-systematicaly
when the script is launched by cron BUT it works systematicaly when the SAME
script is launched manually in a root console.

When the crash occurs the entire shell script is destroyed (there is other
commands and an echo just after the rdiff-backup call that aren't executed).

I have tried to launch the script by hand (./script.sh in a root console)
after adding a '-v9' option in the script, and for the first time,
rdiff-backup has crashed when launched manually.

The first error visible in the log is :

Sending back exception [Errno 11] Resource temporarily unavailable of type
exceptions.IOError: 
  File "/var/lib/python-support/python2.4/rdiff_backup/connection.py", line
335, in answer_request
    result = apply(eval(request.function_string), argument_list)

There is full log attached.

An additional information : the partition incriminated is local ext3, without
other mount points (no NFS and so on). This partition is the biggest to
backup, around 25GiB and the backup take more than 45 min in general.

An idea may be : if ssh connexion isn't used for a long period (when build
filelist ?) maybe sshd destroy it. 



    _______________________________________________________

File Attachments:


-------------------------------------------------------
Date: mercredi 23.07.2008 à 08:40  Name: backup.log.bz2  Size: 16 ko   By:
ludolpif

<http://savannah.nongnu.org/bugs/download.php?file_id=16176>

    _______________________________________________________

Reply to this item at:

  <http://savannah.nongnu.org/bugs/?23915>

_______________________________________________
  Message posté via/par Savannah
  http://savannah.nongnu.org/