Backups to Synology NAS

[dburris]

Hi,

I am trying to configure a Synology DS209+II as a Linux backup destination for
Veeam backups.  I was able to get scp loaded and working on the Synology,
however Veeam is failing to connect to the Synology as a backup server.  When I try
to access the Synology from Veeam Backup & FastSCP 3.1.1, I get this error:

Timed out waiting for operation "(cd /tmp && perl
veeam_soap926aa553-44f5-4b9c-997b-3ef139d65f6d.pl -d -c -l
lib926aa553-44f5-4b9c-997b-3ef139d65f6d -e
/tmp/veeam_error926aa553-44f5-997b-3ef139d65f6d 2>>
/tmp/veeam_error926aa553-44f5-4b9c-997b-3ef139d65f6d) || cat
/tmp/veeam_error926aa553-44f5-4b9c-997b-3ef139d65f6d 2>&1"

The error file in /tmp on the Synology shows:
tar: lib: Not found in archive
Can't open perl script "veeam_soap2d566c77-f611-4038-8b6a-c12ceff49828.pl": No
such file or directory

Perl is in the path and working on the Synology. Is there a way to make this work?

Thanks,
Dave

[Gostev]

Dave, yes it should be possible to make it work. Actually, recently I've got the same error recently on minimal install of the latest Debian Lenny (which also had perl, but still did not work), and it did not work until I have installed build-essential package (I needed it for something else), but this package also made it working. Obviously this package contains tons of stuff, and most is not needed for us... by the way, what I noticed is that it installs some additional perl modules too.

I will consult with our Linux devs tomorrow and update this thread.

[Gostev]

Dave, this looks to be tar issue, probably some "exotic" tar version on your device that does not understand standard command line keys.

Please copy veeam_soap.tar manually to /tmp and run the following command:

Code: Select all

tar -x -f veeam_soap.tar lib veeam_soap.pl

Also please check what is the tar version:

Code: Select all

tar --version

Please let me know the results, thank you.

[dburris]

I copied the veeam_soap.tar to /tmp on the Synology and ran:

tar -x -f veeam_soap.tar lib veeam_soap.pl
gives this error:
tar: lib: Not found in archive

lib is in the tar archive. I went ahead and untarred the whole archive. The lib directory and veeam_soap.pl were extracted to /tmp successfully.

tar --version gives me a usage error. I don't see a switch that would give us the version.
/bin/tar is a symbolic link to busybox.

I still cannot connect to the Synology in Veeam. Gives the same timeout error.
The error in /tmp is:
tar: lib: Not found in archive
Can't open perl script "veeam_soap6a13bf8f-caa9-4bb0-ab13-b94afd14c912.pl": No such file or directory

The name of the veeam_soap.pl file name changes each time I try to connect.

[Gostev]

Definitely tar version issue, since even --version command fails.

Could you please send output of the following commands:

Code: Select all

tar –usage
tar –help
ls –la `which tar` 

[dburris]

DiskStation> tar -usage
tar: invalid option -- u
BusyBox v1.1.0 (2009.09.03-17:10+0000) multi-call binary

Usage: tar -[czjxtvO] [-f TARFILE] [-C DIR] [FILE(s)] ...

DiskStation> ls -la `which tar`
lrwxrwxrwx 1 root root 7 Oct 9 14:28 /bin/tar -> busybox

[Gostev]

Dave, our Linux dev has reviewed the BusyBox 1.1 sources (3 years old release btw).
It appears that tar version used there is outdated and does not work as our code expects tar to work.

Could you please check a few more things:

Does this command unpack the tar correctly (please pay special attention to lib directory):

Code: Select all

tar -x –v -f veeam_soap.tar

What are the other busybox applets NAS distro has installed - you can use this command to show all:

Code: Select all

/bin/busybox

[Gostev]

Dave, we did some further research on Synology NAS, and it appears that its hardware is not x86-based (PowerPC). So even if we can make the above work, our file transfer agent will not work there

[dburris]

Is there any way to get a PowerPC version of the transfer agent?

[Gostev]

I will talk to devs about this once v4 is out

[dburris]

Gostev,

Rather than trying to add the Synology as a backup destination using the "Add Server" button, I ended up just backing up to a shared folder on the Synology. That seems to work well. So instead of using scp, Veeam is using SMB correct? Assuming the Synology (or a Linux server) could be added using the "Add Server" button, are there advantages to backing up via scp rather than SMB?

Thanks for all your help!

Dave

[Gostev]

Dave - yes, this way it backups to share using SMB. This is most typical way of intergrating Veeam Backup with network attached storage devices.

There is two thing which can be considered benefits if you backup to Linux servers:

1. If you are using "Best" compression option, actual compression will be offloaded to the target Linux server from Veeam Backup proxy. Such "distributed processing" may help in some scenarios, for instance you can now run multiple jobs in parallel if writing to different Linux targets. Without this, single job with "Best" compression will load most modern CPUs on Veeam Backup server to 100%.

If you are using lower compression level, then it does not matter because "Optimal" (light) compression is always done by Veeam Backup server as part of network traffic compression, so in this case target server just stores the incoming data without repackaging it.

Anyway, in your specific case, the target server does not have powerful CPU, so you can disregard this bullet alltogether.

2. If you use "full" ESX (not ESXi) and "Network" backup mode, this gives you direct-to-target backups. In other words, Veeam Backup agent running in ESX service console will send data direct to you target Linux box, instead of data going through the Veeam Backup proxy.

[adamswann]

We're considering purchasing a Synology device to place in one of our offices to receive backups from Veeam B&R from our data center.

Our ideal scenario is that we would let Veeam B&R connect directly to the Synology using SSH and run the Linux backup agent directly on the Synology device.

Is this setup known to work? From my research, at least as of a few years ago, the Synology devices include Perl 5.8.6. Are there any other requirements we should be aware of? Are there any Synology models (we're looking at the DS412+ but wouldn't mind going with one of the cheaper models) that might handle this better than others?

Thanks,

Adam Swann

[Vitaliy S.]

Hi Adam,

If Synology NAS satisifes these requirements, then you will be able to add it as a Linux repository:
1. The server must have Perl installed
2. The server supports password authentication
3. The root password should not contain certain symbols (such as space)

Thanks!

[adamswann]

Earlier posts on this thread (which looks like it dates back to 2009) indicate that x86 is also a requirement. Is that still the case? Or will an ARM-based Linux work as long as it meets the three requirements you listed?

For reference, Synology has a mix of ARM and x86 models:

http://forum.synology.com/wiki/index.ph ... y_NAS_have

The "+" models generally (though apparently with some exceptions) run x86 (Intel Atom) CPUs.

I'm probably going to pick up at DS412+ to try out -- just hoping someone could chime in and say, "yes I've made this work!" and save me some surprises.

We're using a Synology at our datacenter with Veeam with no problems, but in that case, it's just exposed as an SMB share.

Thanks,

Adam Swann

[dellock6]

Hi Adam, I've worked with Qnap appliances but I think the requirements are the same, Veeam has binaries for X86-based linux, and it handles NASes as linux servers. Thus I'm pretty sure you would need at least the Atom-based models.
Also, Atom models are to be preferred for performance reasons, on Qnap at least ARM based models are too slow to run decent backups on it (unless you need to save only a bunch of VMs...)

Luca.

[Vitaliy S.]

Earlier posts on this thread (which looks like it dates back to 2009) indicate that x86 is also a requirement. Is that still the case?

I haven't checked it, but I believe it is still the case.

[craig.anderson]

I've got this working on my RS812+ Synologies (which use x86 architecture) but it does not work on on my D213's (the repository can be setup but the backup jobs fail)

One issue I have though is that the capacity/free space always reports as 8388608.0 TB (quite obviously this is not correct)

It doesn't hinder my backups but it would be nice if Veeam were aware oif the remaining backup capacity.

Anybody have any ideas on this?


Also, is there anywhere that I can get more detailed information on what *exactly* the difference is between a transport-agent-enabled repository and a CIFS or NFS share?

Obviously the transport agent takes care of some of the processing so that the proxy does not have to (which I presume results in less cpu load on the proxy and less network traffic between the proxy and repository) but I'm unsure of exactly what it does and how that might benefit or hinder certain jobs/scenarios (incrementals, synthetic fulls, reverse incremental, low bandwidth WAN connections). I've already poured through the online documentation and done some forum searching, but it's still not clear.

If I could understand this better I'd have a better case for using higher quality repositories (read: more expensive) and probably be able to design overall better backup solutions.

I'm also hoping that the upcoming Veeam 7 Backup Copy jobs and WAN Acceleration features can be taken advantage of by a Synology running the transport agent.. I think that would really open up some possibilities.

Thanks!

[Vitaliy S.]

Hi Craig,

One issue I have though is that the capacity/free space always reports as 8388608.0 TB (quite obviously this is not correct)

Do you see these wrong numbers reported by the appliance itself of by the Veeam backup console?

Also, is there anywhere that I can get more detailed information on what *exactly* the difference is between a transport-agent-enabled repository and a CIFS or NFS share?

Obviously the transport agent takes care of some of the processing so that the proxy does not have to (which I presume results in less cpu load on the proxy and less network traffic between the proxy and repository) but I'm unsure of exactly what it does and how that might benefit or hinder certain jobs/scenarios (incrementals, synthetic fulls, reverse incremental, low bandwidth WAN connections). I've already poured through the online documentation and done some forum searching, but it's still not clear.

If I could understand this better I'd have a better case for using higher quality repositories (read: more expensive) and probably be able to design overall better backup solutions.

If you decide to use CIFS shares as a backup target, then it will increase your data traffic between proxy and the NAS box, moreover the rebuild traffic (synthetic full) will not be kept locally and will pushed back and forth over the network which will decrease the overall job performance. Additionally, throttling rules (if you decide to use them) will not be applied as they require our run-time on the NAS box for communication. Finally, with all network glitches that might happen during the backup job run network shares cannot be considered as the best possible backup target.

Hope this helps!

[ChrisRomer]

Vitaliy,

Very new Veeam customer here that has a follow up question to this discussion. I created an iSCSI target on our Synology DS713+ (connected to a DX513) for our Hyper-V host. This works well for our small environment, although Synology broke their LUN backup in DSM 4.3 and that isn't slated for fix until DSM 5. I assume this iSCSI would share those same concerns as what you stated with a CIFS share? If that is the case, and Synology keeps dragging their feet on a LUN backup fix, I will be attempting a transport-agent-enabled repository as suggested by Craig. Please point me to any articles that might help with that.

Much appreciated,
Chris

Thanks,
Chris

[Vitaliy S.]

Hi Chris,

I assume this iSCSI would share those same concerns as what you stated with a CIFS share?

Not really, iSCSI is the preferred way of adding these devices to your backup repository. All the concerns that relate to CIFS shares are not applicable here, since Veeam data movers will be installed on the Windows Server with the attached NAS box. This box will act like a local drive to your repository, so no issues here.

Thanks!

[rcarstens]

craig.anderson, you said you were able to get this working on an 812+? What did you have to do to get it working? I have an 812+ and keep getting the timeout error.

[avit]

I've got this working on my RS812+ Synologies (which use x86 architecture) but it does not work on on my D213's (the repository can be setup but the backup jobs fail)

One issue I have though is that the capacity/free space always reports as 8388608.0 TB (quite obviously this is not correct)

It doesn't hinder my backups but it would be nice if Veeam were aware oif the remaining backup capacity.

Anybody have any ideas on this?

I have a support case open with Veeam at the moment related to this. I have a Synology DS1513+ and using Veeam Backup & Replication treating the Synology NAS as a Linux server so I can use vPower NFS etc. I also noticed that it was reporting 8388608.0TB (which, just out of interest, if you expand it out to bytes, is equivalent to one binary zero followed by 63 binary 1's, which would be the largest 64-bit signed-integer number possible...). Anyway, during troubleshooting with the Veeam support guy, he noticed that the logfile Util.VolumesHostDiscover.log had an entry that said the following (sorry about the lack of line breaks, the log lines are really really long, I've added a carriage return between lines):

Code: Select all

[01.02.2014 04:32:33] <01> Error    /bin/df -P -x vmfs  returned non-zero code\n - /bin/df: invalid option -- x\nBusyBox v1.16.1 (2013-11-06 05:22:56 CST) multi-call binary.\n\nUsage: df [-Pkmh] [FILESYSTEM]...\n\nPrint filesystem usage statistics\n\nOptions:\n	-P	POSIX output format\n	-k	1024-byte blocks (default)\n	-m	1M-byte blocks\n	-h	Human readable (e.g. 1K 243M 2G)\n   at Veeam.Backup.EsxManager.XmlCommandBuilder.ValidateFeedback(String parData)

[01.02.2014 04:32:33] <01> Error       at Veeam.Backup.EsxManager.XmlCommandBuilder.ValidateFeedbackNoErrorXmlException(String parData)

What is happening here is that the VolumesHostDiscover module of Veeam Backup is trying to issue the command "/bin/df -P -x vmfs", but the BusyBox implementation of Linux used on my Synology NAS (and probably all of them) does not support the -x command, which tells it to exclude from the df listing all filesystems of type 'vmfs'. What Veeam sees is the "usage" reply from the df command, which obviously doesn't make sense to it.

I tweaked the perl script that sends this df with the -x option, and simply removed the text "-x vmfs" then saved the perl script and tar'd up the veeam_soap.tar package and put it back in the Veeam folder on my server. Now if I Rescan that repository, I see the proper size reported. There was no risk to me tweaking the script because my NAS doesn't have any VMFS filesystems on it anyway. I can only guess that when VolumesHostDiscover (the module that populates dialog boxes with your repository information, size, free space left etc) receives the wrong response back from BusyBox's df, it results in that spurious terabyte disk size.

We still haven't resolved the main issue I'm having with backing up to the NAS as a Linux server, and in actual fact this issue with df had absolutely nothing to do with the problem... I'm glad I looked into it, because I was all ready to accept Veeam Support's diagnosis that it was a problem with the df command. But having put a workaround in place by tweaking the script to remove "-x vmfs", I have proven that this isn't the cause.

[veremin]

Thank you for sharing such a valuable information; much appreciated.

Also, I'm wondering what particular issues you have while trying to backup to the said NAS device added as Linux server.

[avit]

Thank you for sharing such a valuable information; much appreciated.

Also, I'm wondering what particular issues you have while trying to backup to the said NAS device added as Linux server.

Hi,

The problem I'm having is this (steps to reproduce the problem every time):

1. Erase all restore points for my "Daily Backup" job, from the repository (Synology NAS acting as Linux server)
2. Run a full backup (reverse-incremental 'mode') - this works fine, resulting in one full set of backups for all 11 of my VMs
3. The next time (and every subsequent time) the incremental runs, it will fail with the error message:
"Error: Storage file '/volume1/homes/admin/veeam_backups/Daily_Backup_1/Daily_Backup2014-02-19T210937.vbk' is missing from host 'MYV2-NAS-001'. If you are using initial replica seeding, please follow the instruction from the readme.txt file in the chosen seeding location. If you changed replica destination in the replication job settings, please make sure you have moved all replica files to the new destination correctly."

Oddly, the storage file in question definitely does exist. I can ssh from a terminal window, or navigate to it using Synology DSM's File Explorer, and the file is there for sure. I've checked that the user account specified in the Veeam settings has adequate rights (the user is 'admin' and as you can see, the backup is stored in admin's home folder structure, so I'd be surprised if a file/folder created by admin was unable to be located or read by admin...) - I also tried using the root account for Veeam credentials temporarily, just to be sure that there were no permissions problems. At this stage I'm waiting for a Veeam tech to get back to me - he was available on 3 consecutive days to help me with the case, but unfortunately I was totally slammed with some crisis stuff at work. Hopefully we'll manage to find a time to work together and solve this, and I'll be sure to report back. In the meantime, I'm always open to suggestions on further things to try...

[avit]

Another interesting fact about this, which I hadn't previously realised: every week on Sunday we run a full backup, and that job just ran completely fine to the NAS. So it's literally something to do with the difference between the incrementals and the full backup - something Veeam does differently. So I now have 2 sets of restore points saved on that NAS (if this was working correctly, I'd have 8 -- the two full backups plus the 6 incrementals between the two full backups).

[Vitaliy S.]

2. Run a full backup (reverse-incremental 'mode') - this works fine, resulting in one full set of backups for all 11 of my VMs
3. The next time (and every subsequent time) the incremental runs, it will fail with the error message:

If you run an active full backup, then your job creates a new backup chain and should no longer care about previous points, at least your backup job should not fail.

="avit" Erase all restore points for my "Daily Backup" job, from the repository (Synology NAS acting as Linux server)

Also can you please clarify why do you this or this was done just on testing purposes?

[jim3cantos]

What is happening here is that the VolumesHostDiscover module of Veeam Backup is trying to issue the command "/bin/df -P -x vmfs", but the BusyBox implementation of Linux used on my Synology NAS (and probably all of them) does not support the -x command, which tells it to exclude from the df listing all filesystems of type 'vmfs'. What Veeam sees is the "usage" reply from the df command, which obviously doesn't make sense to it.

I tweaked the perl script that sends this df with the -x option, and simply removed the text "-x vmfs" then saved the perl script and tar'd up the veeam_soap.tar package and put it back in the Veeam folder on my server. Now if I Rescan that repository, I see the proper size reported. There was no risk to me tweaking the script because my NAS doesn't have any VMFS filesystems on it anyway. I can only guess that when VolumesHostDiscover (the module that populates dialog boxes with your repository information, size, free space left etc) receives the wrong response back from BusyBox's df, it results in that spurious terabyte disk size.

We still haven't resolved the main issue I'm having with backing up to the NAS as a Linux server, and in actual fact this issue with df had absolutely nothing to do with the problem... I'm glad I looked into it, because I was all ready to accept Veeam Support's diagnosis that it was a problem with the df command. But having put a workaround in place by tweaking the script to remove "-x vmfs", I have proven that this isn't the cause.

Can you elaborate a little more how to make this change? I have found veeam_soap.tar in veeam server but I can't see any perl file inside with the command (df) to tweak.

thanks.

[jim3cantos]

update: I found the file to tweak (mount.pm) in veeam_soap.tar. I "untared" it, updated with worpad and "tar'd" again with 7-zip. I copied back to original folder and tried to resyncronize repository but it fails (after a while) with following error:

Code: Select all

[i]21/03/2014 10:30:30          Starting synchronization of backup repositories for all backup jobs
21/03/2014 10:30:31          Found 1 backup repositories
21/03/2014 10:32:12 Error    Processing backup repositories
21/03/2014 10:32:12 Error    Failed to synchronize backup repository backup copy linux direct Error: Timed out waiting for operation "(cd /tmp && perl veeam_soap3e961f6f-3bf5-46bc-87e9-ced1c41d2384.pl -d -c -l lib3e961f6f-3bf5-46bc-87e9-ced1c41d2384 -e /tmp/veeam_error3e961f6f-3bf5-46bc-87e9-ced1c41d2384 2>> /tmp/veeam_error3e961f6f-3bf5-46bc-87e9-ced1c41d2384) || cat /tmp/veeam_error3e961f6f-3bf5-46bc-87e9-ced1c41d2384 2>&1", timeout: 100000 ms
21/03/2014 10:32:12 Error    Failed to perform backup repositories synchronization[/i] 

restored original veeam_soap.tar and syncronitation works ok....enough testing for today....

[dellock6]

Be careful with wordpad and other windows editors, maybe they changed some line break....