Adding FreeBSD server as Backup Repository

[Klaus_F]

Hi,

I try to add a Freebsd Server as Backup Repository, I choose the Linux server, enter IP address, I enter user name, password, set to elevate to root and set add to sudoer and root password.

I get auth failed.

I have tried to use putty from this a virtual Windows 2008R2 server where Veeam runs from, then there is no problems with connecting and auth.

I have tried to raise log level on the SSH server on my FreeBSD box, I can see that a connections is make and it failes with out any further messages.

I run 6.0.0.153 (64bit)

Any suggestions to solve this problem?

[Gostev]

Hi Klaus, and welcome to our forums. Please be sure to read the sticky FAQ topics when you get a minute.
FreeBSD is not supported for backup repositories as per System Requirements. (Windows and Linux only).
Thanks!

[Klaus_F]

Hi Gostev,

Thanks for a quick reply.

It is just sometimes that only Linux is mentioned and it works on both Linux and FreeBSD, as the protocols are standard.

Have you any plans to support FreeBSD, as this is a very stable and used in as NAS server like the FreeNAS?

Do you have any preferred linux distributions?

[Gostev]

Hi Klaus, there are no "preferred" Linux distributions, we support any flavor of Linux.
We have no plans to add native support for FreeBSD based on market share considerations.

[poulpreben]

Slight progress on this

Code: Select all

[10.09.2012 21:15:27] < 16384> {
[10.09.2012 21:15:27] < 16384>   Linux agent.
[10.09.2012 21:15:27] < 16384>   Agent version:6.1.0.181
[10.09.2012 21:15:27] < 16384>   PID:17847
[10.09.2012 21:15:27] < 16384>   uname
[10.09.2012 21:15:27] < 16384>     sysname : Linux
[10.09.2012 21:15:27] < 16384>     release : 2.6.16
[10.09.2012 21:15:27] < 16384>     version : FreeBSD 8.2-RELEASE-p2 #0: Fri Sep  2 04:18:07 CEST 2011
[10.09.2012 21:15:27] < 16384>     machine : i686
[10.09.2012 21:15:27] < 16384>   cpuinfo
[10.09.2012 21:15:27] < 16384>   meminfo
[10.09.2012 21:15:27] < 16384> }

It does however fail on a POSIX error before writing data to the VBK file. I will keep digging!

Edit: Included the error message below.

Code: Select all

[10.09.2012 21:15:30] < 65541> cli| Waiting for the next command.
[10.09.2012 21:15:30] < 65541> cli| Next client command: [connectByIPs].
[10.09.2012 21:15:30] < 65541> cli|       :>  10.11.12.5
[10.09.2012 21:15:30] < 65541> cli|       :>  192.168.26.17
[10.09.2012 21:15:30] < 65541> cli|       :>  VPTESTVBR1
[10.09.2012 21:15:30] < 65541> cli|       :>  .
[10.09.2012 21:15:30] < 65541> cli|       :>  2501
[10.09.2012 21:15:30] < 65541> cli|       :>
[10.09.2012 21:15:30] < 65541> cli|       :>  {8b4186a3-f2f1-4486-8ae1-3a85e43256d5}
[10.09.2012 21:15:30] < 65541> cli|       :>  {0f5ae9e7-6d60-415c-98d7-e99f4d053998}
[10.09.2012 21:15:30] < 65541> cli| Connecting to the server agent with the following parameters: [10.11.12.5;192.168.26.17;VPTESTVBR1,2501].
[10.09.2012 21:15:30] < 65541> cli|   Trying to connect to the endpoint [10.11.12.5:2501]
[10.09.2012 21:15:30] < 65541> cli|   Connection status: system:0 ( Success ).
[10.09.2012 21:15:30] < 65541> cli| Connecting to the server agent with the following parameters: [10.11.12.5;192.168.26.17;VPTESTVBR1,2501]. Failed.
[10.09.2012 21:15:30] < 65541> cli| Removing stdin/stdout redirector: '143428080'.
[10.09.2012 21:15:30] < 65541>      ERR |Protocol not available
[10.09.2012 21:15:30] < 65541>      >>  |POSIX: Unable to update socket keep-alive settings. Error code: [%1].
[10.09.2012 21:15:30] < 65541>      >>  |--tr:Client failed to process the command. Command: [connectByIPs].
[10.09.2012 21:15:30] < 65541> net| Connection closed...

Regardless, this will be an unsupported solution, as it requires detection of the Linux emulation kernel modules etc. which will probably be too much work for development considering the demand for FreeBSD support. I am just a ZFS lover, which is why I am interested in seeing if I can make it work.

[Ua3prq]

It is just sometimes that only Linux is mentioned and it works on both Linux and FreeBSD, as the protocols are standard.
Have you any plans to support FreeBSD, as this is a very stable and used in as NAS server like the FreeNAS?
Do you have any preferred linux distributions?

Hi ALL!
We use FreeNAS in production for a long time, no problem (CIFS share). In VM also...

[NTmatter]

A bit of quick testing shows that this is works on FreeBSD 10. It's not exactly bug-free (reports 8 exabytes of space), and the repository setup wizards likely won't do the job correctly, and this configuration may void your warranty and set fire to all three copies of your data.

That being said, setup is fairly straightforward if you're already familiar with FreeBSD setup.

Base System:
Start with a fresh install of FreeBSD 10 from the DVD media. Initial testing shows that the ZFS Root is quick and usable. The extra packages are not directly required, so Kernel, Sources, Games, 32-Bit Libraries, and Ports do not need to be installed. Set up networking as desired, and ensure that SSHD is configured to run.

When prompted to set up user accounts, create your usual admin account and an extra account for Veeam. Note that the Veeam account will require passwordless sudo to root, so ensure it has a strong password.

Now complete the setup and reboot into the live system.

Dependencies:
Veeam Backup Repositories require SSH (installed by default), Perl, and Sudo. Linux compatibility is surprisingly not required to get things running. Become root to install Perl and Sudo:

Code: Select all

su -
pkg install perl5 sudo

Set up passwordless sudo for the Veeam user:

Code: Select all

echo "veeam ALL=(root) NOPASSWD: ALL" >> /usr/local/etc/sudoers.d/veeam

Enable Password authentication in /etc/ssh/sshd_config:

Code: Select all

PasswordAuthentication yes

Restart SSHD to pick up the change:

Code: Select all

service sshd restart

Veeam Setup:
This is effectively a normal Linux server and repository setup.

Start the Add Server wizard, and select the "Linux" option. Provide the DNS name or IP to the FreeBSD server, and an appropriate description.

Select or create credentials to match the Veeam account on the FreeBSD system. Ensure that "Elevate specified account to root" is checked. Ensure that "Add account to the sudoers file" is not checked, as we have manually set up passwordless sudo to root for the account.

Now create a Linux backup repository. Select the server that was just created. Note that the Populate button will throw up an error stating "Unable to retrieve drive info" -- this does not seem to be a significant issue. Advance to the next step, and select a folder to house your repository. Note that this populate button will report 8388608.0 TB of space -- this is also not a significant issue for testing. You may also opt to align backup data blocks if you have enough memory to support ZFS deduplication for your full backup set.

Finish off the vPower NFS setup, and complete the setup.

You should now be able to run backups directly to the datastore without resorting to a Samba install.

[lp@albersdruck.de]

I am just a ZFS lover, which is why I am interested in seeing if I can make it work.

You could also try zfsonlinux.org.
I do use this in a production environment and it works just fine - no deduplication though, as that server has not enough RAM (only 24 GB). Adding this server as a repository to Veeam works as expected, no problems so far.
zfs send | receive is a wonderful backup copy mechanism for off-site backups.

[poulpreben]

Hi Lars,

In my personal home lab, I am currently using ZFS on Linux for hosting the Veeam repository. One major limitation I have found with using ZFS with a Veeam workload is when running reversed incremental.

As you probably know, ZFS is using a copy-on-write transactional model, meaning that active blocks are never overwritten in-place, but instead a new block is allocated, and the data is written there. This method is diminishing the whole point of running reversed incremental, where we are trying to retain the latest point in time as a full backup. On ZFS, effectively this full backup (VBK file) will be fragmented across the filesystem and the performance penalty is really significant.

However, if you are running regular forward incremental and leave deduplication enabled for the zvol, the space savings are great (similar to what can be achieved with Windows Server 2012 deduplication). And by adding a few SSD drives for ZIL and L2ARC in front of regular 7.2k RPM SATA drives, I have seen amazing results.

As much as I love ZFS (zfs send | get over SSH <3), there are still a lot of issues with it. Especially the fact that with current zpool versions, performance will drop to 20-30% if a zpool is filled beyond 90-95%, and it will never recover in performance until the zpool is destroyed and recreated.

Personally, I cannot think of any filesystem I would rather use for my backups, but it will take some time before I will recommend it to Average Joe.

Cheers,
Preben

[bmalicoat]

I just wanted to second the zfsonlinux.org comment. I'm using it on Ubuntu with no issues in production.
(although we use iscsi targets so the backup drive looks like a local NTFS volume on the backup server, and the linux fileserver is actually a VM)

[bia2014]

Good morning,

in my setup every night i perform a reverse incremental of Vmware infrastructures, target destinations are LAN(gigabit) samba shares with ZFS(onlinux) underlying filesystem. I feel they're just taking too much time to complete, so i would like to know if there are specific suggestions Veeam software recommend in order to achieve maximum performances with this kind of setup. Cut and paste from logs just to explain the order of magnitude.

Start time 20:00:08
Total size 968,0 GB
Backup size 315,1 GB
End time 07:24:31 (+1)
Data read 65,8 GB
Dedupe 1,7x
Duration 11:24:22
Transferred 15,7 GB
Compression 1,9x

Without being too specific, given that multiple variables may be on the ground, there any tweaks on both sides (Veeam/filesystem) that you may suggest as a starting point from your experience? I've heard about improvements adjusting block size on target destination filesystem, but just rumors, nothing "official".

Thank you

[PTide]

Hi,

Your topic has been merged. Please take a look at this post.

Thank you.

[datagen24]

Here is a potential workaround to get the benefit of ZFS but abstract it using Jails

In illumos you can do much the same thing using a branded zone

what we are doing is creating a jail with Centos 6 user land I am trying it in my lab tonight and we shall see

good instructions here, they are for an older release (10.2) of freebsd but the concepts are the same in the new release
https://bluehatrecord.wordpress.com/201 ... ebsd-10-2/

[poulpreben]

That is definitely a workaround. I am no Linux expert, but is it not emulating a relatively old kernel version? 2.6.18 has over 400 CVEs listed, so I personally think I would like to avoid using anything like this in production. It seems to be the same kernel version for FreeBSD 11.0-RELEASE.

The docker/freebsd-compat branch could also be an idea, but relies on same Linux ABI as your suggestion.

If you really want to go down the rabbithole of virtualizing Linux on top of FreeBSD for ZFS "support", you may want to take a look at bhyve instead. It is included natively with FreeBSD 10.0, and the performance is really great. You can get a full CentOS, which can be patched and updated, and you can export ZFS volumes to the VMs directly.

[adapterer]

We use FreeNAS but found the merge performance godawful over SMB no matter what we tried.

The answer is use ZVOL + iSCSI (file based iSCSI uses up too much space and can't be reclaimed for thin provisioning)