Hello everyone,

I have run into a problem similar to the one Andrea and Chris are
describing. I want to share what worked for me and a couple of tips to
avoid problems.

In my case, I had 100TB of files with incorrect mtime and ctime stamps. I
corrected those stamps from a mirror of those files (using a touch
script). Bacula saw all the times change and recommended backing up the
entire 100TB again. I knew the hashes of those files shouldn't have
changed, so I experimented with the 'accurate' 'o' option flag.

FYI, this is a strange NAS situation, where I only have access to the NAS
appliance via SMBv3 (works better than you'd think), and permissions are
fairly flat, and controlled by the NAS, not by my SMB client. So I only
really care about backing up the files, and their modified dates.


>From the bacula 13.x + manual, fileset 'accurate' option:
========================
o Save only the metadata of a file when possible (Available from version
13.0.0) .
The ’o’ option should be used in conjunction with one of the signature
checking
options (1, 2, 3, or 5). When the option is specified, the signature is
computed only
for files that have one of the other accurate options specified triggering
a backup of
the file (for example an inode change, a permission change, etc...). In
cases where
only the file’s metadata has changed (ie: the signature is identical), only
the file’s
attributes will be backed up. If the file’s data has been changed (hence a
different
singature), the file will be backed up in the usual way
========================


In the end, I found:

1. I ultimately used 'accurate = Mo3'. 'M' means "check mtime and ctime".
Anyone using this option should obviously select whichever hash algorithm
they specified in 'signature=xyz' (or probably 5 for the default, md5).

2. I was previously using 'accurate = yes' for this fileset. When I changed
the option to 'accurate = Mo3', this did not constitute a fileset change
that would trigger a new full backup (so you don't have to turn on the
potentially dangerous 'ignore fileset changes = yes' to avoid provoking a
new full backup).

3. I used the 'estimate' command to verify whether bacula was going to do
what I wanted, before running the job. Once I changed to use the accurate
'o' flag, the estimate processes took a very long time, because they had to
hash all the files with changed metadata. I originally thought I'd use
something like 'accurate = pinso3'* (this didn't work!)*. The bacula manual
said:
==================
accurate=<options> The options letters specified are used when running a
Backup
Level=Incremental/Differential in Accurate mode.
*The options letters are the sameas in the verify= option below with the
addition of the following options:*
==================
I believed that 'accurate = pinso3' would work. It did not.

The bacula director did not throw an error when I reloaded the config, BUT
an estimate said I would have 100TB of files to back up, which tells me it
did not work correctly. It did take several days to estimate, so I think it
hashed the files.

I changed the option to 'accurate = Mo3', and ran the estimate again. This
time, bacula thought it would have to back up 30TB of files. This is the
result I expected. I didn't check before making the mtime changes to my
files, but it's pretty plausible that we could have had 30TB of new files
on disk waiting for a backup.

4. Please note that in bconsole, the estimate command defaults to
level=full. You need to specify level=incremental in order to see
comparisons of how much data would be backed up in an incremental backup.

5. Eric posted some text from the manual about the 'accurate' 'o' feature.
Towards the end he said "This feature will be available in the next 17.0
that is still cooking in our lab." 'Accurate = Mo3' is definitely working
in bacula 13.x. I think right now the only options supported by this
feature are 'M' or maybe 'mc' before 'o(hash algorithm). I think he means
that bacula CE 17.x will support more expanded options like ’pinug2o’, not
that the feature doesn't work at all in bacula CE right now.

I hope this helps.

Regards,
Robert Gerber
402-237-8692
[email protected]
_______________________________________________
Bacula-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-users

Reply via email to