Hello Justin, Thanks for your response and feedback.
For me the statement (a) is correct, but let's try to look at this problem from another side. Is it possible to see some outputs like the verify job log and also fileset and the verify job configuration? Maybe together we will be able to find something there that could cause the problem. Best regards, Marcin Haba (gani) On Fri, 9 Sept 2022 at 17:14, Justin Case <jus7inc...@gmail.com> wrote: > Thanks Marcin, I actually read that some time ago. In my opinion it is > kinda weird that the SD reads the files from the volume and sends the whole > files to the FD, just for hashing. Thats sounds really inefficient. > > The point where I think that the documentation - and your article - are > mistaken is about level=data with accurate=yes: > > There can be only one of these statements right: > > either > (a) file size and checksum (these are actually hashes, not checksums, but > hey) are taken from the backup volume for comparison, > > or > (b) file size and checksum (these are actually hashes, not checksums, but > hey) are taken from the disk/file system for comparison, > > for Data with Accurate the documentation and your article claim (a). > > If (a) was true, it were inconsistent with the outcome of my little > experiment (see previous email, below). > Statement (b), however, seems to be consistent with the experiment, at > least for the comparisons only happening when accurate=yes. > > Bottom line: with accurate=yes, Level=Data, the last part in your flow > diagram is not actually behaving like VolumeToCatalog, but like > DiskToCatalog. > > Do you see why I am confused here? Where have I gone wrong? Is my > experiment not working or is the code not behaving as documented? > > Best, > J/C > > > On 9. Sep 2022, at 16:43, Marcin Haba <ganius...@gmail.com> wrote: > > Hello Justin, > > In verify jobs Bacula components have different tasks to do. File daemon > is capable to read file attributes from the filesystem, compute checksums > and also compare file size and checksums. Storage daemon is capable to read > data and metadata from volumes. Director is capable to compare file > attributes in the Catalog with file attributes and checksums sent from File > daemon. Depending on which verify job level you use, different Bacula > components are involved in the verify process. > > Some time ago I prepared an article about verify jobs. I described there > how each of the verify job level works including flow diagrams that show > what is going inside Bacula components when verify jobs are running. There > is also information which Bacula components are used to which verify job > level. You can find the article here: > > > https://bacularis.app/news/10/36/Bacula-verify-job-and-a-new-Bacularis-verify-wizard/d,Bacularis%20news%20details > > I hope you find it useful. Good luck. > > Best regards, > Marcin Haba (gani) > > On Fri, 9 Sept 2022 at 15:51, Justin Case <jus7inc...@gmail.com> wrote: > >> I guess I don’t understand how verify jobs work at level Data with >> Accurate=yes. >> >> When scheduled, the verify level=data and accurate=yes job runs well >> after the backup job, but returns with differences, but not showing which >> files have differences. >> When run again manually, the same happens. >> >> When running the backup job (incremental) again, and after that the >> verify job, I will also get differences, but now it shows which files are >> causing them! >> That are literally a few hundred! But all of them are some caches, >> database files of macOS, Firefox, etc. >> >> This is kinda puzzling me as I thought that Level=Data does not access >> the client FD, but with Accurate=yes it seems to do that. >> I thought this compares the data in the volumes to the metadata in the >> catalog (as opposed to to the files as seen by the client FD). >> If that was the case, how could those differences occur? >> And how could they be so numerous, even though I am running the verify >> job right after the backup job? >> >> So I ran a Level=VolumeToCatalog verify and no differences found. >> Then I ran a Level=DiskToCatalog verify and voila lots and lots of >> differences. >> >> Could it be that Level=Data with Accurate=yes does NOT encompass >> VolumeToCatalog (as per the documentation), but really encompasses >> DiskToCatalog (which would be uncalled for, really)? >> >> That still doesn’t answer why the scheduled verify jobs didn’t display >> the files that had the differences as opposed to the manually run verify >> jobs, which did show the files with differences. >> >> Really, what is going on here? Is nobody else using this and wondering? >> Did I get it all wrong? >> >> > On 9. Sep 2022, at 02:25, Justin Case <jus7inc...@gmail.com> wrote: >> > >> > Finally 8 of 15 verify jobs returned with differences. I did not have >> this before when I was only using level Data and not Accurate. >> > How can I investigate teh root cause of this? >> > >> > What seems counter intuitive to me is that all verify jobs are marked >> as OK, also those that returned with differences. >> > I guess this means the verify job ran completely and has no relation to >> potential differences being found. >> > >> >> On 9. Sep 2022, at 01:23, Justin Case <jus7inc...@gmail.com> wrote: >> >> >> >> Hello all, >> >> a disk backup verify with level Data and with Accurate just returned >> with >> >> FD termination status: Verify differences >> >> >> >> However, the job info and the email that was sent about it does not >> contain any further information on what file is exhibiting differences. Is >> it possible to investigate further? I would be interested which files are >> showing difference. I did not find any console commands that seemed >> suitable for that. >> >> >> >> Another thing that caught my attention is this: >> >> bacula-dir JobId 2753: Sending Accurate information to the FD. >> >> >> >> I was thinking that Data level verify does not involve the client FD. >> Does Accurate cause Data level verify to involve the client FD? If that is >> the case it would explain the verify differences, as that client is macOS >> and has a lot of files that change frequently. >> >> >> >> Thanks for your interest, >> >> J/C >> >> >> >> >> >> >> >> >> > >> >> >> >> _______________________________________________ >> Bacula-users mailing list >> Bacula-users@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/bacula-users >> > > > -- > "Greater love hath no man than this, that a man lay down his life for his > friends." Jesus Christ > > "Większej miłości nikt nie ma nad tę, jak gdy kto życie swoje kładzie za > przyjaciół swoich." Jezus Chrystus > > > -- "Greater love hath no man than this, that a man lay down his life for his friends." Jesus Christ "Większej miłości nikt nie ma nad tę, jak gdy kto życie swoje kładzie za przyjaciół swoich." Jezus Chrystus
_______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
