[Bacula-users] Odd error with -2.5-b2

Hi

I've successfully updated my Linux hosts to 2.5b2 and testing has been
mostly rather successful. The new accurate backup mode appears to work
very well even on my Cyrus mail spools (several hundred thousand files).

However, I'm having issues with my NT profiles directory. It's a
dedicated ext3 filesystem, /ntprofiles, with POSIX ACLs and user xattrs
enabled on it.

It's failing with (sorry for >quoted text; Thunderbird is too stupid to
do non-wrapped text with normal paste):

> 26-Mar 23:17 backup-dir JobId 372: Start Backup JobId 372, 
> Job=AccessNtProfiles.2009-03-26_23.10.00.39
> 26-Mar 23:17 backup-dir JobId 372: Using Device "FileStorage_HomeDir"
> 26-Mar 23:15 access.localnet-fd JobId 372: DIR and FD clocks differ by -113 
> seconds, FD automatically compensating.
> 26-Mar 23:17 backup-dir JobId 372: Sending Accurate information.
> 26-Mar 23:17 backup-sd JobId 372: Volume "HomeDirIncr-0400" previously 
> written, moving to end of data.
> 26-Mar 23:17 backup-sd JobId 372: Ready to append to end of Volume 
> "HomeDirIncr-0400" size=311720709
> 26-Mar 23:17 backup-sd JobId 372: Spooling data ...
> 26-Mar 23:17 backup-sd JobId 372: Job write elapsed time = 00:00:21, Transfer 
> rate = 2.720 M bytes/second
> 26-Mar 23:17 backup-sd JobId 372: Committing spooled data to Volume 
> "HomeDirIncr-0400". Despooling 57,259,094 bytes ...
> 26-Mar 23:17 backup-sd JobId 372: Despooling elapsed time = 00:00:01, 
> Transfer rate = 57.25 M bytes/second
> 26-Mar 23:17 backup-sd JobId 372: Sending spooled attrs to the Director. 
> Despooling 395,586 bytes ...
> 26-Mar 23:15 access.localnet-fd JobId 372: Fatal error: backup.c:944 Network 
> send error to SD. ERR=Connection reset by peer
> 26-Mar 23:15 access.localnet-fd JobId 372: Fatal error: backup.c:1048 Network 
> send error to SD. ERR=Connection reset by peer
> 26-Mar 23:15 access.localnet-fd JobId 372: Fatal error: backup.c:1048 Network 
> send error to SD. ERR=Connection reset by peer

These messages continue to repeat thousands of times. The email
notification is 10Mb.

(I've fixed the NTPd issues causing the clock drift since then).

The same host, access.localnet, has no problems talking to the sd
(backup-sd on backup.localnet) when backing up the Cyrus mail spools on
the same host, which are also set to Accurate=yes.

I haven't re-run with debug enabled yet, but I was looking for a few
ideas on where to start tracking this down. The only thing I can think
of so far is that this is the only fileset that backs up POSIX ACLs and
user xattrs. I'll try turning user xattrs off and will report more info
later, but given the beta status I thought it best to bring this up
promptly.

Here are some relevant config snippets:

Job {
  Name = "AccessNtProfiles"
  JobDefs = "DefaultJob"
  Schedule = "HomeDirSchedule"
  Storage = "File_HomeDir"
  Client = access.localnet-fd
  FileSet = "AccessNtProfiles"
  Accurate = "yes"
}
FileSet {
  Name = AccessNtProfiles
  Include {
    Options {
      signature = MD5
      compression = gzip
      onefs = yes
      # Samba uses ACLs on these files
      aclsupport = yes
      # Needed because Samba stores hidden etc in user xattrs
      xattrsupport = yes
    }
    File = /ntprofiles
  }
}

--
Craig Ringer

------------------------------------------------------------------------------
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users