On Tuesday, March 19, 2019 at 5:04:26 PM UTC+1, Gaetan SLG wrote:
> Hi everyone,
> 
> We are planing to implement AI jobs for one of our projects and we decided to 
> make tests before on a running infrastructure (Version: 18.2.6).
> 
> The problem is AI jobs are backing up less files than original incr job. To 
> do our test we copied a normal job :
> 
> 
> *list job=denver-always-incremental
> Using Catalog "MyCatalog"
> +--------+---------------------------+--------+---------------------+------+-------+----------+---------------+-----------+
> | jobid  | name                      | client | starttime           | type | 
> level | jobfiles | jobbytes      | jobstatus |
> +--------+---------------------------+--------+---------------------+------+-------+----------+---------------+-----------+
> | 21,197 | denver-always-incremental | denver | 2019-03-15 15:03:10 | B    | 
> F     |  147,998 | 2,079,207,003 | T         |
> | 21,217 | denver-always-incremental | denver | 2019-03-16 01:51:21 | B    | 
> D     |      341 |    32,224,492 | T         |
> | 21,254 | denver-always-incremental | denver | 2019-03-18 01:15:39 | B    | 
> I     |       17 |     1,524,336 | T         |
> | 21,290 | denver-always-incremental | denver | 2019-03-19 01:15:34 | B    | 
> I     |       20 |     4,419,477 | T         |
> +--------+---------------------------+--------+---------------------+------+-------+----------+---------------+-----------+
> *list job=denver
> ....
> 
> | 21,216 | denver | denver | 2019-03-16 01:50:56 | B    | D     |      341 |  
>   32,223,670 | T         |
> | 21,253 | denver | denver | 2019-03-18 01:14:34 | B    | I     |    1,291 |  
>  184,462,873 | T         |
> | 21,289 | denver | denver | 2019-03-19 01:14:43 | B    | I     |    2,413 |  
>  132,643,290 | T         |
> +--------+--------+--------+---------------------+------+-------+----------+---------------+-----------+
> 
> 
> As you can see, starting from same base (same differential job on 2019-03-16) 
> does not provides same results.
> 
> 
> * File set is same
> * "estimate" provides same results. Only Incr jobs are "buggy"
> 
> Here is a copy of our config files :
> 
> job
> {
>   Name = "denver"
>   FileSet = "denver"
>   Schedule = "WeeklyCycle"
>   Client = "denver"
>   JobDefs = "DefaultJob"
>   Write bootstrap = "/var/lib/bareos/bstraps/denver.bsr"
>   Accurate = yes
> }
> 
> Job
> {
>   Name = "denver-always-incremental"
>   FileSet = "denver"
>   Schedule = "WeeklyCycleAlwaysIncremental"
>   Client = "denver"
>   JobDefs = "DefaultJob"
>   Write bootstrap = "/var/lib/bareos/bstraps/denver-always-incremental.bsr"
>  
>     # Always incremental settings 
>     AlwaysIncremental = yes 
>     AlwaysIncrementalJobRetention = 7 days 
>     Always Incremental Keep Number = 6
>     
>     Accurate = yes 
>  
>     Pool = AI-Incremental 
>     Full Backup Pool = AI-Consolidated 
> }
> 
>  
> Job { 
>     Name = "denver-always-incremental-consolidate" 
>     Type = "Consolidate" 
>     Accurate = "yes" 
>     JobDefs = "DefaultJob" 
> }
>  
> 
> FileSet
> {
>   Name = "denver"
>   Ignore FileSet Changes = yes
>   Include {
>     Options {
>       signature = MD5
>       compression = gzip
>       noatime = yes
>     }
>     File = /
>     File = /boot
>     File = /home
>     File = /var/log
>   }
>   Exclude {
>     File = /proc
>     File = /sys
>     File = /tmp
>     File = /.journal
>     File = /.fsck
>   }
> }
> 
> 
> 
> 
> JobDefs {
>   Name = "DefaultJob"
>   Type = Backup
>   Level = Full 
>   Client = baloo
>   FileSet = "Full Set"
>   Schedule = "NONE"
>   Storage = File 
>   Messages = Standard
>   Pool = Month
>   Priority = 10 
>   Accurate = yes
> }
> 
> 
> Do is it a bad configuration or it could be a bug ?
> 
> Thank you !


Hi,

I think I've found the error. This looks like implementation error. This looks 
related to the accurate implementation and the situation where 2 incrementals 
(different jobs but on same client, sharing same accurate informations) are run 
sequentially on the same Client.
As a try, I've inverted the backup order : Launching the AI before the original 
Incremental, and... The result is also inverted  !

So, I created a "testmodif1.txt" file, then launching AI Job then Incr job with 
tracelog enabled and the result is : 

root@pve-support:~# grep testmo /var/lib/bareos/pve-support-fd.trace.*
/var/lib/bareos/pve-support-fd.trace.ai:pve-support-fd (100): 
filed/accurate.cc:186-21381 accurate /root/testmodif1.txt (not found)
/var/lib/bareos/pve-support-fd.trace.incr:pve-support-fd (100): 
filed/accurate_htable.cc:87-21382 add fname=</root/testmodif1.txt> lstat=P0A 
DAAKQ IGk B A A A A BAA A Bck55J Bck55J Bck55J A A d delta_seq=0 chksum=
/var/lib/bareos/pve-support-fd.trace.incr:pve-support-fd (100): 
filed/accurate.cc:101-21382 lookup </root/testmodif1.txt> ok

As you can see, the file is not backed up on the second incr job.


Again, modified the testmodif1.txt file and launching AI then Incr job  : 

root@pve-support:~# grep testmod /var/lib/bareos/pve-support-fd.trace*
/var/lib/bareos/pve-support-fd.trace.ai:pve-support-fd (100): 
filed/accurate_htable.cc:87-21383 add fname=</root/testmodif1.txt> lstat=P0A 
DAAKQ IGk B A A A A BAA A Bck55J Bck55J Bck55J A A d delta_seq=0 chksum=
/var/lib/bareos/pve-support-fd.trace.ai:pve-support-fd (100): 
filed/accurate.cc:101-21383 lookup </root/testmodif1.txt> ok
/var/lib/bareos/pve-support-fd.trace.ai:pve-support-fd (99): 
filed/accurate.cc:276-21383 /root/testmodif1.txt      st_mtime differs
/var/lib/bareos/pve-support-fd.trace.ai:pve-support-fd (100): 
findlib/bfile.cc:1101-21383 bopen: fname /root/testmodif1.txt, flags 00000000, 
mode 0000, rdev 0
/var/lib/bareos/pve-support-fd.trace.incr:pve-support-fd (100): 
filed/accurate_htable.cc:87-21384 add fname=</root/testmodif1.txt> lstat=P0A 
DAAKW IGk B A A A F BAA I Bck6Aw Bck6Aw Bck6Aw A A d delta_seq=0 chksum=
/var/lib/bareos/pve-support-fd.trace.incr:pve-support-fd (100): 
filed/accurate.cc:101-21384 lookup </root/testmodif1.txt> ok


Then as you can see, as soon as the file have been backed up a first time, it 
will never be backed up again even if the job is not the same. As a result : 
Inconcistent backups

I know it should not happen with a proper jobs setup but in some cases, it 
could.

-- 
You received this message because you are subscribed to the Google Groups 
"bareos-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to