From: Holger P. <wb...@pa...> - 2008-10-19 20:50:47
|
Hi, Jeffrey J. Kosowsky wrote on 2008-10-19 14:58:15 -0400 [Re: [BackupPC-users] Incremental dumps hanging with 'Can't get?rsync digests' & 'Can't call method "isCached"']: > Nils Breunese (Lemonbit) wrote at about 19:50:35 +0200 on Sunday, October 19, 2008: > > Jeffrey J. Kosowsky wrote: > > > When running an incremental backup on my Linux system, BackupPC > > > repeatedly hangs when backing up the mythconverg mysql database for > > > mythtv. > > > [...] > > > Can't get rsync digests from > > > /var/lib/BackupPC/pc/consult/3/f%2f/fvar/flib/fmysql/fmythconverg/ > > > frecord.MYD > > > (err=-2, name=var/lib/mysql/mythconverg/record.MYD) > > > Can't call method "isCached" on an undefined value at > > > /usr/share/BackupPC/lib/BackupPC/Xfer/RsyncFileIO.pm line 165. a quick glance at the code reveals nothing (to me). I'll have a closer look later. > > > The most disturbing thing to me is that BackupPC seems to be so > > > fragile in that if it encounters a problem, it seems to just hang > > > there (seemingly) forever without sending any error messages I'm sure Craig will happily accept your patch. > Now interestingly, I did a new FULL backup and now subsequent > INCREMENTALS seem to be working. But I will keep monitoring the > situation. > > I imagine that either something (maybe the attrib file for that > directory?) was written correctly on the initial full backup (or was > later corrupted though not likely since I'm running RAID and the > backup is fresh). This then caused BackupPC to crash when it hit the > malformed part of the full backup. 1.) RAID does *nothing* to prevent on-disk file corruption. RAID lets you recover from detected read error (which would otherwise be signalled as read errors, not silently ignored). 2.) Did you run the script I recently posted [1] to check the attrib files? I'd like to think the time I spent coding it was not completely wasted. If we actually found corrupt attrib files, that would be a start to making BackupPC less "fragile", as you put it. A further sanity check I should probably add to the script is a test for the existance of files listed in the attrib file, but that's not as simple as it sounds, because they may be in a preceeding backup of lower level rather than the same one the attrib file is in. Which makes me wonder: could the "backups" file be corrupted rather than an attrib file? Regards, Holger [1] http://sourceforge.net/mailarchive/message.php?msg_name=20081007041523.GB14536%40gratch.parplies.de |