Ok thanks for the infos,
Il will look for this script too
-Message d'origine-
De : Bill Arlofski via Bacula-users
Envoyé : lundi 4 mars 2024 17:41
À : bacula-users@lists.sourceforge.net
Objet : Re: [Bacula-users] CRC ERROR on volume what possible to do
On 3/4/24 09:00, Lionel PLASSE wrote:
> Hello,
>
> I have this error while reading a volume for virtualfull consolidating
> job
>
> Error: block_util.c:521 Volume data error at 0:0!
> Block checksum mismatch in block=11755301 len=64512: calc=dfb5486f
> blk=9eafdf1a
>
> It seems the volume file is definitively corrupted, nothing is
> possible for this block I think . ( The error occurs twice for the
> same job at the same
> block)
>
> But ,
> is this possible to continue reading the volume by bypassing or
> "marking" the error and proceed with consolidating the remaining data?
> Can we blacklist the block in error (along with the corresponding
> files) to complete the consolidation job, even if the result will be
> an incomplete fileset?
> Or Bacula definitively kill the job in error. I don't recall seeing an
> option to bypass I/O errors .
>
> What to do in such kind of hardware I/O problems.
Hello Lionel,
I have no idea if this would work, but it may be possible to start the SD with
the `-p` (Proceed despite I/O errors), then try the restore. I have never tried
this, and would typically revert to using the low-level `bextract` tool, which
also has the `-p` command line option.
If starting the SD with `-p`:
# sudo -u bacula /path/to/bacula-sd -p -f(just start the SD in foreground
with ignore errors etc)
... And performing a VFull does not get you a good* virtual full, you may have
to use `bextract` against the volumes used in the last Full/VirtualFull to
restore the data from the volumes.
*In this sentence, the word "good" is relative. I mean, with hardware I/O
errors, the data recovered during the VFull will surely be missing data... The
same thing goes for using bextract set to ignore I/O errors.
Personally, unless this were a critical restore situation, and I were just
trying to "get what I can" back, I would abandon the last Full/VirtualFull and
perform a new, good real Full immediately since you have surely lost some data
in your backup chain due to this hardware issue.
I'd also recommend implementing Verify jobs in some manner (ie: Automatically
restore and/or Verify (level=data) critical jobs when they finish with a
RunsWhen = after script, or implement some Admin job that pseudo-randomly picks
a recent, Good backup job and performs a restore and/or Verify job against it)
Here's an all-in-one™, overkill™ example script that I wrote as a
proof-of-concept a while ago which performs a restore, then all three Verify
levels against a backup job when it completes. You can pick and choose parts
of this that you need and abandon the rest. :)
https://github.com/waa/AutoRestoreAndVerify
Hope this helps,
Bill
--
Bill Arlofski
w...@protonmail.com
smime.p7s
Description: S/MIME cryptographic signature
___
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users