Bacula-users

[Bacula-users] Where are file signatures generated? Can bad memory on SD system cause data corruption?

2010-01-20 15:34:07
Subject: [Bacula-users] Where are file signatures generated? Can bad memory on SD system cause data corruption?
From: Michel Meyers <steltek AT tcnnet DOT com>
To: bacula-users <bacula-users AT lists.sourceforge DOT net>
Date: Wed, 20 Jan 2010 20:56:50 +0100
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

Not sure if somebody on the users list can answer this, but with a
fileset that has this:

FileSet {
  Name = "Backup"
  Include {
   Options { signature=MD5; }
...


1. Where are the signatures actually generated?
2. Are the signatures checked when a restore is done?


I'm asking because I recently had a weird issue with Bacula and am
wondering whether my diagnosis is correct.

Backstory: I recently inherited an Exabyte 110L LTO2 Loader (IBM
Ultrium-TD2 drive) and PCI-E SCSI Controller (Adaptec ASC-29320ALP
U320). My storage 'server' doesn't have a PCI-E slot, so I hooked all
this up to my newer 'desktop' PC and installed an SD there (both run
Bacula 3.0.3 on Debian Testing with a 2.6.30 kernel). To test the setup
I started a backup with a few bigger files (mostly CD images) from the
server and then restored and compared them to the originals. The restore
process didn't complain about the files but to my surprise, they didn't
match. A closer inspection showed that all of the files only differed in
a single byte.

A few weeks later my desktop PC went haywire, which I eventually
diagnosed as faulty memory (segfaults and gpfs in Linux and both
memtest86+ and Windows Memory Diagnostic found issues with 3 of the 4
RAM bars). Knowing that this PC has had sporadic issues previously (with
Windows Vista dumping to blue screen and complaining about the RAM when
it came back) I replaced all of the memory modules and figured I'd
retest Bacula again. It turns out that this backup and restore worked
flawlessly (same amount of files backed up and test files, but it did
use a different tape than in the first test).

Hence my question: How could the file difference slip past the restore
process and is it possible that faulty memory on the SD can cause silent
data corruption? Or is that not possible and the fault actually lies
with either the tape, the drive or the controller/controller cable?

- - Michel

P.S. the relevant part from my SD configuration:

Autochanger {
  Name = Exabyte110L
  Device = Ultrium-TD2
  Changer Command = "/etc/bacula/mtx-changer %c %o %S %a %d"
  Changer Device = /dev/changer
}

Device {
  Name = Ultrium-TD2
  Drive Index = 0
  Media Type = LTO-2
  Archive Device = /dev/nst0
  AutomaticMount = yes;               # when device opened, read it
  AlwaysOpen = yes;
  Spool Directory = /tapedump
  RemovableMedia = yes;
  RandomAccess = no;
  AutoChanger = yes
  Maximum File Size = 3GB
  Maximum Spool Size = 10000m
}

The /tapedump spool directory is on a RAID1 array.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (MingW32)

iEYEARECAAYFAktXYAIACgkQ2Vs+MkscAyWqQgCfbMdYFNE7RjkbP3JLoDJn4ASq
8z8AoJLkML8L3cr+sF5kKFwbdoRLIAO1
=kthU
-----END PGP SIGNATURE-----

------------------------------------------------------------------------------
Throughout its 18-year history, RSA Conference consistently attracts the
world's best and brightest in the field, creating opportunities for Conference
attendees to learn about information security's most important issues through
interactions with peers, luminaries and emerging and established companies.
http://p.sf.net/sfu/rsaconf-dev2dev
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>