Disk to Catalog Verification

We will now add the next step towards “end-to-end” integrity checking of your backups – the DiskToCatalog verification. We first complete our verification job of the /etc directory on the client, and then add the needed job to verify the source files against the catalog by adding the following job resources to our configuration:

Job {
  Name = Verify-RH1-etc-Job-Volume
  Type = Verify
  Verify Job = RH1-config
  Level = VolumeToCatalog
  Client = s-rh1-fd
  JobDefs = DefaultJob
}

Job {
  Name = Verify-RH1-etc-Job-Disk
  Type = Verify
  Verify Job = RH1-config
  Level = DiskToCatalog
  Client = s-rh1-fd
  JobDefs = DefaultJob
}

The configuration should be rather self-explanatory by now – Verify-RH1-etc-Job-Volume ensures we can run the verification of the latest backup on media any time, and Verify-RH1-etc-Job-Disk allows us to run the corresponding verification against the source files.

If we run the CatalogToDisk verification, we get a result as shown in the figure below.

main-dir JobId 28: Verifying against JobId=26 Job=RHi-config.2010-10-17_01.25.44 _49
main-dir JobId 28: Start Verify JobId=28 Level=DiskToCatalog Job=Verify-RHi—etc-Job-Disk.2010-10-17 _02.21.23_57
main-dir JobId 28: Bacula Enterprise main-dir 4.0.4 (04Sep10):
  Build:                  x86_64-unknown-linux-gnu suse 11.1
  JobId:                  28
  Job:                    Verify-RH1-etc-Job-Disk.2010-10-17 02.21.23 _57
  FileSet:                etc-fs
  Verify Level:           DiskToCatalog
  Client:                 s-rhi-fd
  Verify JobId:           26
  Verify Job:             RH1-config
  Start time:             17-0ct-2010 02:2
  End time:               17-0ct-2010 02:2
  Files Examined:         2,216
  Non-fatal FD errors:    0
  FD termination status:  OK
  Termination:            Verify OK

main-dir JobId 24481: Begin pruning Jobs older than 3 hours.
main-dir JobId 24481: No Jobs found to prune.
main-dir JobId 24481: Begin pruning Jobs.
main-dir JobId 24481: Pruned Files from 2 Jobs for client s-rhi-fd from catalog.
main-dir JobId 24481: End auto prune.

We will now change the verified metadata of one of the files verified, add a new file, and remove a file that existed at the time of the latest backup:

[root@lsb-254 etc]# pwd
/etc
[root@lsb-254 etc]# touch resolv.conf
[root@lsb-254 etc]# touch NEWFILE
[root@lsb-254 etc]# mv dnsmasq.conf /tmp/

The job report from a subsequent Verify-RH1-etc-Job-Disk job clearly indicates the differences we created (actually, it does more – it implies that /etc/resolv.conf and /etc/sysconfig/networking/profiles/default/resolv.conf are hard links to the same data).

main-dir JobId 25: Verifying against JobId=24 Job=RH1-config.2010-10-17_01.25.44_49
main-dir JobId 25: Start Verify JobId=25 Level=DiskToCatalog
                   Job=Verify-RH1-etc-Job-Disk.2010-10-17_02.31.25_58
main-dir JobId 25: File: /etc/sysconfig/networking/profiles/default/resolv.conf
main-dir JobId 25: st_ctime differs
main-dir JobId 25: st_mtime differs
main-dir JobId 25: File: /etc/resolv.conf
main-dir JobId 25: st_ctime differs
main-dir JobId 25: st_mtime differs
main-dir JobId 25: New file: /etc/NEWFILE
main-dir JobId 25: File: /etc/
main-dir JobId 25: st_ctime differs
main-dir JobId 25: st_mtime differs
main-dir JobId 25: Warning: The following files are in the Catalog but not on disk:
main-dir JobId 25: /etc/dnsmasq. conf
main-dir JobId 25: Bacula Enterprise main-dir 4.0.4 (04Sep10):
  Build:                  x86_64-unknown-linux-gnu suse 11.1
  JobId:                  25
  Job:                    Verify-RH1-etc-Job-Disk.2010-10-17 02.31.25_58
  FileSet:                etc-fs
  Verify Level:           DiskToCatalog
  Client:                 s-rhi-fd
  Verify JobId:           24
  Verify Job:             RH1-config
  Start time:             17-0ct-2010 02:31:27
  End time:               17-0ct-2010 02:31:40
  Files Examined:         2,216
  Non-fatal FD errors:    0
  FD termination status:  OK
  Termination:            Verify Differences

This verification just checked “real” metadata and does not prove that file contents changes will be noticed. However, a job report containing lines like

File: /etc/hosts
      st_size  differ. Cat: 187 File: 222
      st_ctime differs
      st_mtime differs
      SHA1 digest differs. File=CRPld0oeAo4ugQ2/HCV0/JJe/EY Cat=rZUB6O05vQ3P28HUvi9B/O7917Y

indicates just those sort of changes. (With the File Set used in the example, the size and file time changes are also noted, of course.)

We have seen how easy it is to set up CatalogToDisk verification; it gets even easier once the configuration is generated automatically, as only very minimal information in the job resource needs to be adjusted. We have also seen that changes to the files on disk are noticed and reported, including the actual information that has changed.