Incremental backup do same (static) files.

49 views
Skip to first unread message

vol...@gmail.com

unread,
Apr 27, 2019, 10:54:15 AM4/27/19
to bareos-users
Hi, All

After implementing AlwaysInc schema, I've payed attention that my incremental backup are quite large. Probably it was also true for classic Full-Diff-Inc schema or not, I am not sure. Inspecting files included in Incremental backup, I can see same static unchanged files backed up every day. I thought about MD5 hash collision, then I've replaced the signature to be SHA256. After Full backup, a next Incremental brings me almost same list of backed up files. The noatime directive also not helped.

What is strategy of selecting files to be backed up ?
If "accurate=pins5" in fileset is works in AlwaysInc mode , or its own "accurate=yes" in jobdef breaks selection ?

Also, I saw still open bug https://bugs.bareos.org/view.php?id=907 related to Windows. I see the same problem on linux.

Bareos Version is 17.2.4
My Fileset is:
FileSet {
Name = "HANA-001"
Description = "Backup local ext2,3,4 and xfs FS"
Include {
Options {
signature=SHA256
onefs=Yes
compression=GZIP
accurate=pins5
sparse=yes
noatime=yes
# keepatime=yes
}
File = "\\|sh -c \"mount | egrep 'ext[2-4]|xfs' | awk '{print \$3}' \""
Exclude Dir Containing = .nobareos
}
Exclude {
File = /var/lib/bareos
File = /var/lib/bareos/storage
File = /proc
File = /sys
File = /tmp
File = /var/tmp
File = /dev
File = /.journal
File = /.fsck
}
}

My jobdef is
JobDefs {
Accurate = yes
Always Incremental = yes
Always Incremental Job Retention = 30 days
Always Incremental Keep Number = 6
# Always Incremental Max Full Age = 31 days
Client = bareos-fd
FileSet = "SelfTest" # selftest fileset
Full Backup Pool = Full
Incremental Backup Pool = Incremental
Level = Incremental
Messages = Standard
Name = "AlwaysInc"
Next Pool= "Full"
Pool = "Incremental"
Priority = 10
Schedule = "AlwaysInc"
Storage = File
Type = Backup
}

Andreas Rogge

unread,
Apr 29, 2019, 3:55:21 AM4/29/19
to bareos...@googlegroups.com
Am 27.04.19 um 16:54 schrieb vol...@gmail.com:
> After implementing AlwaysInc schema, I've payed attention that my incremental backup are quite large. Probably it was also true for classic Full-Diff-Inc schema or not, I am not sure. Inspecting files included in Incremental backup, I can see same static unchanged files backed up every day. I thought about MD5 hash collision, then I've replaced the signature to be SHA256. After Full backup, a next Incremental brings me almost same list of backed up files. The noatime directive also not helped.

You may have hit a bug here. We have heard of issues like that before
but could never pinpoint what exactly happens here and why.
As we cannot reproduce what is broken here, we will need somebody who
can reproduce and test in his envrionment.
Are you able and willing to do this? You might need to run patched
versions of the file-daemon and director for that.

> What is strategy of selecting files to be backed up ?
> If "accurate=pins5" in fileset is works in AlwaysInc mode , or its own "accurate=yes" in jobdef breaks selection ?
I did not understand this. Can you rephrase it please?

> Also, I saw still open bug https://bugs.bareos.org/view.php?id=907 related to Windows. I see the same problem on linux.
While the bug has been filed for a Windows client this will probably
happen on every type of system.

Best Regards,
Andreas
--
Andreas Rogge andrea...@bareos.com
Bareos GmbH & Co. KG Phone: +49 221-630693-86
http://www.bareos.com

Sitz der Gesellschaft: Köln | Amtsgericht Köln: HRA 29646
Komplementär: Bareos Verwaltungs-GmbH
Geschäftsführer: S. Dühr, M. Außendorf, J. Steffens, Philipp Storz

signature.asc

Oleg Volkov

unread,
Apr 29, 2019, 4:26:23 AM4/29/19
to Andreas Rogge, bareos...@googlegroups.com
On Mon, 29 Apr 2019 at 10:55, Andreas Rogge <andrea...@bareos.com> wrote:
Am 27.04.19 um 16:54 schrieb vol...@gmail.com:
> After implementing AlwaysInc schema, I've payed attention that my incremental backup are quite large. Probably it was also true for classic Full-Diff-Inc schema or not, I am not sure. Inspecting files included in Incremental backup, I can see same static unchanged files backed up every day. I thought about MD5 hash collision, then I've replaced the signature to be SHA256. After Full backup, a next Incremental brings me almost same list of backed up files. The noatime directive also not helped.

You may have hit a bug here. We have heard of issues like that before
but could never pinpoint what exactly happens here and why.
As we cannot reproduce what is broken here, we will need somebody who
can reproduce and test in his envrionment.
Are you able and willing to do this? You might need to run patched
versions of the file-daemon and director for that.

I though to build new director with version 18, then move clients to it.
Give me a couple of days for that, it could be testing server for beginning.


> What is strategy of selecting files to be backed up ?
> If "accurate=pins5" in fileset is works in AlwaysInc mode , or its own "accurate=yes" in jobdef breaks selection ?
I did not understand this. Can you rephrase it please?

My fileset definitions has "accurate" directive and my jobdef (both posted in original mail) has also "accurate" directive.
Are they conflicts in some way ?

> Also, I saw still open bug https://bugs.bareos.org/view.php?id=907 related to Windows. I see the same problem on linux.
While the bug has been filed for a Windows client this will probably
happen on every type of system.

Best Regards,
Andreas
--
  Andreas Rogge                             andrea...@bareos.com
  Bareos GmbH & Co. KG                      Phone: +49 221-630693-86
  http://www.bareos.com

  Sitz der Gesellschaft: Köln | Amtsgericht Köln: HRA 29646
  Komplementär: Bareos Verwaltungs-GmbH
  Geschäftsführer: S. Dühr, M. Außendorf, J. Steffens, Philipp Storz

 
Oleg Volkov
T: +972-50-7601914
Skype: voleg4u , Viber, WhatsApp.

Andreas Rogge

unread,
Apr 29, 2019, 4:31:31 AM4/29/19
to Oleg Volkov, bareos...@googlegroups.com
Am 29.04.19 um 10:26 schrieb Oleg Volkov:
> I though to build new director with version 18, then move clients to it.
> Give me a couple of days for that, it could be testing server for beginning.
That would be great. If there is an issue, we really need to find and
fix it.

> > What is strategy of selecting files to be backed up ?
> > If "accurate=pins5" in fileset is works in AlwaysInc mode , or its
> own "accurate=yes" in jobdef breaks selection ?
> I did not understand this. Can you rephrase it please?
>
> My fileset definitions has "accurate" directive and my jobdef (both
> posted in original mail) has also "accurate" directive.
> Are they conflicts in some way ?
No they don't conflict.
The accurate in Job/JobDefs enables and disables accurate
The accurate in the fileset's options configures how accurate decides
wether a file needs to be backed up or not.

Best Regards,
signature.asc

Oleg Volkov

unread,
Apr 30, 2019, 10:44:49 AM4/30/19
to Andreas Rogge, bareos...@googlegroups.com
OK, I've sat.
Interesting, that using another server from scratch and higher version of director, I still see same files in AlwaysIncremental backup.
Lets start debug the issue.

Oleg Volkov
T: +972-50-7601914
Skype: voleg4u , Viber, WhatsApp.

Jörg Steffens

unread,
Apr 30, 2019, 11:43:37 AM4/30/19
to bareos...@googlegroups.com
On 30.04.19 at 16:44 wrote Oleg Volkov:
> OK, I've sat.
> Interesting, that using another server from scratch and higher version
> of director, I still see same files in AlwaysIncremental backup.
> Lets start debug the issue.

From your configuration I see:

FileSet {
Name = "HANA-001"
...
signature=SHA256
...
accurate=pins5

The "5" in "pins5" means "compare the MD5 signature" while your are
storing the SHA256 signature.

I wonder if this may be the cause for your problems.

Jörg

--
Jörg Steffens joerg.s...@bareos.com
Bareos GmbH & Co. KG Phone: +49 221 630693-91
http://www.bareos.com Fax: +49 221 630693-10

Sitz der Gesellschaft: Köln | Amtsgericht Köln: HRA 29646
Komplementär: Bareos Verwaltungs-GmbH
Geschäftsführer:
S. Dühr, M. Außendorf, Jörg Steffens, P. Storz

Oleg Volkov

unread,
Apr 30, 2019, 4:31:22 PM4/30/19
to Jörg Steffens, bareos...@googlegroups.com
You are right, it was MD5 in original, thus 5 presents in accurate option.
Then I thought that source of problem could be MD5 hash collision issue, then I've replaced MD5 with SHA256 and forgot to tune pins5.
Now I cannot find in documentation what exactly should be instead of 5 there.
The only documented option is 1 for SHA1 that is also relatively short.
The server under backup is mirror repository with lot of RPMs with similar names.

Anyway I've changes 5 to 1 and SHA256 to SHA1.
The Incremental job tooks same files as before (Of course I've run Full before)
This mean that SHA1 or MD5 not an issue here, SHA1 also has collisions but they are different as MD5.
The list of files in AlwaysInc should change, but it do _same_ files.

Oleg Volkov
T: +972-50-7601914
Skype: voleg4u , Viber, WhatsApp.

--
You received this message because you are subscribed to a topic in the Google Groups "bareos-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bareos-users/PMuqXDqUAWQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to bareos-users...@googlegroups.com.
To post to this group, send email to bareos...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

vol...@gmail.com

unread,
May 2, 2019, 8:15:54 AM5/2/19
to bareos-users
I've run FD in debug mode, like:
# /usr/sbin/bareos-fd -d 99 -f -m > /tmp/debug 2>&1

Inspecting resulting files, found that accurate.c do correct selection.
It picks up 96 really changed files, summary about 100m.

However, the resulting job shows:

gibuy-dir JobId 22: Bareos gibuy-dir 18.2.5 (30Jan19):
Build OS: Linux-4.4.92-6.18-default redhat CentOS Linux release 7.6.1810 (Core)
JobId: 22
Job: fd-sle11sp4.2019-05-02_13.46.07_48
Backup Level: Incremental, since=2019-05-01 22:10:09
Client: "sle11sp4" 17.2.4 (21Sep17) x86_64-suse-linux-gnu,suse,SUSE Linux Enterprise Server 11 (x86_64),SLE_11_SP4,x86_64
FileSet: "HANA-001" 2019-04-30 22:52:10
Pool: "Incremental" (From command line)
Catalog: "MyCatalog" (From Client resource)
Storage: "File" (From Job resource)
Scheduled time: 02-May-2019 13:46:07
Start time: 02-May-2019 13:46:11
End time: 02-May-2019 13:50:27
Elapsed time: 4 mins 16 secs
Priority: 10
FD Files Written: 260
SD Files Written: 260
FD Bytes Written: 4,159,639,508 (4.159 GB)
SD Bytes Written: 4,159,673,113 (4.159 GB)
Rate: 16248.6 KB/s
Software Compression: 87.7 % (gzip)
VSS: no
Encryption: no
Accurate: yes
Volume name(s): Incremental-0010
Volume Session Id: 5
Volume Session Time: 1556710222
Last Volume Bytes: 4,163,353,822 (4.163 GB)
Non-fatal FD errors: 0
SD Errors: 0
FD termination status: OK
SD termination status: OK
Bareos binary info: bareos.org build: Get official binaries and vendor support on bareos.com
Termination: Backup OK

Probably these "extra" files were not sent to FD with accurate data, then FD think they are missing and new, then should be backed up. May be they are sent by DIR, but because of different buffer sizes, just cutted of the accurate data.

As in additional, this happens only in AlwaysInc mode, the classic Full-Diff-Inc works correct.

Andreaas, Jorg, what you think ?

Oleg Volkov

unread,
May 8, 2019, 9:27:04 AM5/8/19
to bareos-users
Hi,
How can I catch accurate data sent by Director to FD ?
This can help to understand where is a problem.

Oleg Volkov
T: +972-50-7601914
Skype: voleg4u , Viber, WhatsApp.

Andreas Rogge

unread,
May 9, 2019, 4:55:45 AM5/9/19
to bareos...@googlegroups.com
Am 08.05.19 um 15:26 schrieb Oleg Volkov:
> Hi,
> How can I catch accurate data sent by Director to FD ?
> This can help to understand where is a problem.
you can enable tracing on the client with the setdebug command on the
director.
With a large enough debug level the client should write the accurate
file list to the dump file.

Best Regards,
Andreas
--
Andreas Rogge andrea...@bareos.com
Bareos GmbH & Co. KG Phone: +49 221-630693-86
http://www.bareos.com

Sitz der Gesellschaft: Köln | Amtsgericht Köln: HRA 29646
Komplementär: Bareos Verwaltungs-GmbH
Geschäftsführer: S. Dühr, M. Außendorf, J. Steffens, Philipp Storz

signature.asc

Oleg Volkov

unread,
May 10, 2019, 12:30:00 PM5/10/19
to Andreas Rogge, bareos-users
Please give an example, I've already tried this without success.

Oleg Volkov
T: +972-50-7601914
Skype: voleg4u , Viber, WhatsApp.

--
You received this message because you are subscribed to a topic in the Google Groups "bareos-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/bareos-users/PMuqXDqUAWQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to bareos-users...@googlegroups.com.
To post to this group, send email to bareos...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages