Le 14/11/2013 23:48, Matias Banchoff a écrit :
> Hi!
>    We are having a problem between a Bacula server version 5.2.5 (SD and 
> Dir) and a Windows client running Bacula-fd 5.2.10.
>
>    The problem is we cannot make a full backup terminate with success. 
> Backup jobs start without any problem, but they end with error after 
> some GBs transfered (Sometimes after 1GBs, 2GBs, 4GBs. It does not 
> appear to be a network shapping problem).
>    When the volume of data to backup is small, the backup job works 
> fine. The problem arises when we backup the whole data.
>    Nobody has complained about connectivity problems with that Windows 
> Server (it's a database server).
>
>    We have tried the following, but none worked:
> - enable and disable VSS.
> - configure heartbeat interval... everywhere! (SD, dir, FD) :-)
> - lower the tcp keepalive time in our Bacula server (The one with the SD 
> and Dir installed) (sysctl net.ipv4.tcp_keepalive_time).
> - disable Checksum Offload in the NIC config (As someone suggested in 
> http://www.bacula.org/manuals/en/install/install/Client_Fi_daemon_Configura.html).
>  
> Our Windows has an HP nic.
> - transfer data from the Windows FD to an SD in the same network the 
> Windows server is, in an attempt to put SD and FD "closer".
>
>    The next step is to downgrade Bacula-fd from 5.2.10 to 5.2.5 and see 
> if that resolves the problem. Any ideas? or any clue?
>
> Notes:
> - Backups are stored in a storage mounted using NFS.
> - The data we are backing up is a single 50GB file.
>
> Thanks!!!!
>
Hi Matias,

You didn't tell what the error message was...

But anyway, this problem reminds me of the one I had some years ago.
I had a Windows client that suddenly always failed on Full backups, most
of the time around 4GB transferred, no problem at all with Incremental,
no problems with Linux Clients which were in the same network,
and no other applications were having network problem on this Windows
server...
So I tried upgrading both side, tuning the Windows Network parameters
(like checksum offload), played with heartbeat and keepalives... no success.

I finally tested with a simple FTP transfert between the Windows client
and my Linux SD => it failed at approximately 4GB to.
So I ended dumping the network conversation at both side and then
compare the dump in wireshark to find out some that at some time, the
packet were not received, and were retransmited until it really fails.
With those elements I asked the network team to analyse all the network
devices involved between FD and SD, and they find out that a switch
interface had a lot of errors.
After changing it : problem solved !

The fact was that only the Windows Client had the problem, and only
Bacula would triggered it out, my guess is that the Linux TCP/IP pile
was a little more robust than the Windows one.
And it always happened in Full mode because the network is generally
more used on Full backups.

/mysysadminlife

So my advice there :
1) test the network connection between the two hosts, with a high rate
tranfert protocol like FTP (or iperf ?).
   I don't know how iperf handles connection losts, but FTP clients will
show you an error and will try to reconnect, you can't miss it if it
happens.
2) analyse all the network elements involved in the route between your 2
hosts.

Just my 2 cents !

-- 
Yann Cézard - administrateur systèmes serveurs
Centre de ressources informatiques    -     http://cri.univ-pau.fr
Université de Pau et des pays de l'Adour -  http://www.univ-pau.fr
bâtiment d'Alembert (anciennement IFR), rue Jules Ferry, 64000 Pau
Téléphone : +33 (0)5 59 40 77 94


------------------------------------------------------------------------------
DreamFactory - Open Source REST & JSON Services for HTML5 & Native Apps
OAuth, Users, Roles, SQL, NoSQL, BLOB Storage and External API Access
Free app hosting. Or install the open source package on any LAMP server.
Sign up and see examples for AngularJS, jQuery, Sencha Touch and Native!
http://pubads.g.doubleclick.net/gampad/clk?id=63469471&iu=/4140/ostg.clktrk
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users