Re: incremental accurate jobs failing with 11.0.3

Hello,

I have seen this issue recently and I have pushed a fix
in the git repository. I will publish a 11.0.4 today.

commit ff91cf7521b2b2eac5e80806b4353fe7b652b8ff
Author: Eric Bollengier <[email protected]>
Date:   Thu May 6 15:15:29 2021 +0200

    Disable fix on bvfs_get_jobids() temporarily


It has not been detected by the regress scripts, I need
to figure why.

Best Regards,
Eric 

On 5/28/21 10:13, Andrea Conti wrote:
> Hello,
> 
> After updating from 11.0.2 to 11.0.3 (gentoo x86_64, postgresql 13) I started having problems with incremental accurate jobs failing with this error:
> 
> 28-May 08:45 blue-dir JobId 96812: Fatal error: Cannot find previous jobids.
> 28-May 08:45 blue-dir JobId 96812: Fatal error: Network error with FD during Backup: ERR=Interrupted system call
> 28-May 08:45 blue-dir JobId 96812: Fatal error: No Job status returned from FD.
> 
> After some digging around, there seems to be a problem with the query looking for the previous Full job:
> 
> 2021-05-28 08:45:04.378 CEST [19174] LOG:  statement: CREATE TABLE btemp396812 AS SELECT JobId, StartTime, EndTime, JobTDate, PurgedFiles, FileSetId FROM Job JOIN FileSet USING (FileSetId) WHERE ClientId = 18 AND Level='F' AND JobStatus IN ('T','W') AND Type='B' AND StartTime<'2021-05-28 08:45:05' AND FileSet.FileSet=(SELECT FileSet FROM FileSet WHERE FileSetId = 49) *AND JobId = 96812*  AND Name = 'infra.disk.backup'  ORDER BY Job.JobTDate DESC LIMIT 1
> 2021-05-28 08:45:04.382 CEST [19174] LOG:  statement: INSERT INTO btemp396812 (JobId, StartTime, EndTime, JobTDate, PurgedFiles) SELECT JobId, StartTime, EndTime, JobTDate, PurgedFiles FROM Job JOIN FileSet USING (FileSetId) WHERE ClientId = 18 AND Level='D' AND JobStatus IN ('T','W') AND Type='B' AND StartTime > (SELECT EndTime FROM btemp396812 ORDER BY EndTime DESC LIMIT 1) AND StartTime < '2021-05-28 08:45:05' AND FileSet.FileSet= (SELECT FileSet FROM FileSet WHERE FileSetId = 49)   AND Name = 'infra.disk.backup'  ORDER BY Job.JobTDate DESC LIMIT 1
> 2021-05-28 08:45:04.383 CEST [19174] LOG:  statement: INSERT INTO btemp396812 (JobId, StartTime, EndTime, JobTDate, PurgedFiles) SELECT JobId, StartTime, EndTime, JobTDate, PurgedFiles FROM Job JOIN FileSet USING (FileSetId) WHERE ClientId = 18 AND Level='I' AND JobStatus IN ('T','W') AND Type='B' AND StartTime > (SELECT EndTime FROM btemp396812 ORDER BY EndTime DESC LIMIT 1) AND StartTime < '2021-05-28 08:45:05' AND FileSet.FileSet= (SELECT FileSet FROM FileSet WHERE FileSetId = 49)   AND Name = 'infra.disk.backup'  ORDER BY Job.JobTDate DESC
> 2021-05-28 08:45:04.383 CEST [19174] LOG:  statement: SELECT JobId FROM btemp396812 ORDER by JobTDate
> 
> i.e. there is a filter on JobID that should definitely not be there (96812 is the id of the *current* job).
> 
> This looks to me like the changes introduced in 44b8d23f6670b70f7292c45d7e33b43b1ac09dad ('fix for #2500') are for some reason being triggered in an unexpected way.
> 
> Is this a bug or am I misunderstanding something? Do you need more information about my setup?
> 
> Thanks in advance,
> 
> andrea
> 
> 
> 
> 


_______________________________________________
Bacula-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-devel