Re: VirtualFull e client availability...

Bill Arlofski via Bacula-users <[email protected]> Thu, 05 Dec 2024 23:39:02 +0000
Newsgroups	gmane.comp.bacula.user
Message-ID	<[email protected]>
On 12/5/24 7:46 AM, Marco Gaiarin wrote:
> 
> I use virtual full backup to 'consolidate' backups on weekend; AFAI've understood,  virtual full backups are a 'sd only' things.

Hello Marco,

You have understood this correctly.


> But if client is not abailable (common for that client, on weekend), also a virtualfull backup fail:

This should not be the case...


And, the first error I see here is `Cancelling duplicate JobId=31257.`
----8<----
> 05-Dec 15:33 lnfbacula-dir JobId 31268: Rescheduled Job FVG-PP-HFA3-1.2024-12-05_15.33.41_23 at 05-Dec-2024 15:33 to re-run in 1800 seconds (05-Dec-2024 16:03).
> 05-Dec 15:33 lnfbacula-dir JobId 31268: Cancelling duplicate JobId=31257.
----8<----

It seems that jobid 31257 is an earlier job than this one. It seems to me that everything after this `canceling duplicate 
job` error could just be a red herring. It is hard to tell if Bacula has decided to cancel this VFull job 31268 or the other 
one. :)

Can you share the co
nfiguration for your `FVG-PP-HFA3-1` job, the pools in question, etc? (might not be necessary, keep 
reading to the end)

I think this, and anything after the first error may be misleading us...
----8<----
> 05-Dec 15:33 lnfbacula-dir JobId 31268: Error: bsockcore.c:285 gethostbyname() for host "850-16268.dyn.pp.lnf.it" failed: ERR=Name or service not known
> 05-Dec 15:33 lnfbacula-dir JobId 31268: Failed to connect to File daemon.
----8<----

This can also be misleading us:
----8<----
> 05-Dec 15:33 lnfbacula-dir JobId 31268: 3904 Job FVG-PP-HFA3-1.2024-12-05_08.00.00_31 not found.
> 05-Dec 15:33 lnfbacula-dir JobId 31268: Error: bsockcore.c:285 gethostbyname() for host "850-16268.dyn.pp.lnf.it" failed: ERR=Name or service not known
> 05-Dec 15:33 lnfbacula-dir JobId 31268: Failed to connect to File daemon.
> 05-Dec 15:33 lnfbacula-dir JobId 31268: 3904 Job FVG-PP-HFA3-1.2024-12-05_08.00.00_31 not found.
----8<----


BUT... this is a very old version of Bacula Direct
or and SD... I would strongly recommend to upgrade:
----8<----
> 05-Dec 15:33 lnfbacula-dir JobId 31268: Error: Bacula lnfbacula-dir 9.4.2 (04Feb19):
----8<----

Client is quite old too, and should be upgraded at some point, but Clients are not contacted during VirtualFull level jobs, 
so I do not think that this is too important for us:
----8<----
>    Client:                 "fvg-pp-hfa3-1-fd" 7.4.4 (28Sep16) Microsoft  (build 9200), 64-bit,Cross-compile,Win64
----8<----


> Seems that hapens on client that areon dynamic DNS, so effectively client does not exist in DNS, not 'does not ping'.
> 
> Apart putting clients on static addresses to they exist, there's something i can do?

Honestly, I think I would push you to upgrade Bacula to a current version (15.0.2). I seem to vaguely recall that there may 
have been an issue about the VirtualFull attempting to resolve the client unnecessarily.

I just ran some VirtualFull jobs in my environment, and I literally (tempor
arily) set a client's to:
----8<----
Address = bogus.fqdn.top
----8<----

...and I see no such DNS resolution errors in my `level=VirtualFull` joblog, and my VFulls for this client completed normally.


P.S. One simple test (without upgrading) you can do is to temporarily set the `Address = localhost` for the Client 
`fvg-pp-hfa3-1-fd` in the Director's configuration, reload the Director's config, and run this job `FVG-PP-HFA3-1` again with 
`level=VirtualFull` and see what happens. Just check to make sure no other of this job is running due to the `canceling 
duplicate` setting you have enabled.


Hope this helps,
Bill

-- 
Bill Arlofski
[email protected]

_______________________________________________
Bacula-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-users
signature.asc (application/pgp-signature, 509 B)