Failed to start client agent on the host...

[TommyTheKid]

Hi there,

I have a customer who is using Veeam B&R 6 patch 3 to backup locally and (hopefully) remote.

We initially backed up to a USB Disk on a "Windows Server" (desktop that had USB3 support), and the handed the USB disk to the offsite storage guy. He loaded it on his NAS, and I can see/access the files when I added a "Linux" server through the Veeam interface. The "Backups" section shows the backups that were copied, so I believe it can "see" them OK.


However, when I try to change the job to use the new repository, it fails with "Failed to start client agent on the host (hostname). Timeout to start agent" ... I have looked at http://www.veeam.com/kb_articles.html/KB1105 ... but we are not using any "special" characters (unless "SPACE" is considered special?). The backup job is simply named "Remote Backup" (as per above)... Here is some output from the Linux server while it was "Reading backup files" after changing the repository from the local USB disk to the remote Linux server.


... so it sortof bothers me that it looks like there should be some perl stuff running from /tmp, but its not there. There is a "binary" there bit its getting "cannot execute binary file" (wth?).

I have it running via CIFS over the WAN right now, but I don't think that is going to work real reliably, especially when it tries to create a synthetic full, so I would like to get the Linux agent part running. We do have a premium, enterprise support agreement, but since I am not sure whether this is the Veeam software or something with the remote Linux server, I thought I'd ask here first.

Thanks in advance,
Tommy

[foggy]

Tommy, when a Linux repository is used, Veeam B&R uses a run-time agent to process backup data locally. This small agent is installed through SSH login and has a perl based wrapper script (so the server must have perl installed). The user account provided while creating the repository should be able to login via SSH and have appropriate permissions to the backup folder.

You can check if you can start the agent manually on the server. If you cannot, then there is something with the server indeed. Otherwise, I would suggest to contact our support directly to investigate.

Also, I've removed log snippets from your post. Please avoid posting them as explained when you click New Topic.

[TommyTheKid]

I have access to the "NAS" host via SSH, and I have set it up as Linux server in Veeam. Veeam can see the files (the entire filesystem) if I expand the server. I believe that to be working. Perl is installed and working (ok I just made a basic "hello world" one liner, but it worked. What versions, libraries, etc are required? Maybe a KB would be good for this, once we detail this? I am thinking that maybe they need to install some "legacy" libc packages or something (gotta love binaries). It works fine to my linux "desktop" (Ubuntu 11.10), but not to their server (CentOS 5.7). I am thinking it may be related to a missing package. Sorry about missing the "no logs" part, I was trying to put as much information in as I could. (I will go back and read that)

**EDIT**
Ah, I was supposed to go down the support path first, sorry . I will file a ticket, but unfortunately, the "license" (support agreement) is tied to my colleague's account, so I need to work with him to get it filed. (that is why I went down the community forum path)

Some more things I have learned since filing the original case...

1. Backup jobs with spaces DO work (tested it here on my "lab" setup)

2. Even over LAN, using a "Linux Server" appears to be *much faster* than using SMB/CIFS, so now I want this to work even more

3. Energy Drinks ROCK!

[foggy]

Have you tried running our agent manually from /tmp? Do you possibly have it mounted with "noexec"?

[Vitaliy S.]

Btw, what is the make and model of your NAS box used as a backup repository?

[TommyTheKid]

Good Idea foggy, but /tmp is just part of /, ext3, rw. I also tried copying /bin/bash to /tmp just to verify, and it does run from there. I did try to run the agent directly and got something to the effect of can not execute binary file (they cut that portion out of my post)

We will have two repositories...
LOCAL: DroboFS (with 4 x 2TB SATA) ... which I wouldn't wish upon my enemies (too slow)
REMOTE: Bit Refinery - Denver (Service provider) ... it looks like a pretty standard CentOS 5.7 Linux box with a RAID

I can think of why it wouldn't run there.

[Gostev]

So if you are trying to do this with your DroboFS box, this will not work of course, because DroboFS is not built on x86 platform.

[TommyTheKid]

DroboFS is local, this is a problem with the remote Linux host (NAS). The Linux system is a CentOS 5.7 x64 box, pretty standard.

Support Ticket: 5175893

More details... no ports are firewalled on the remote host, I have tested using "cat /etc/hosts | nc -l 2500" then trying to telnet to port 2500 from the veeam host.

[maccyber]

[merged]

Hi guys

I'm trying to use FreeNAS (x86_64) as a backup repository.
Anyone out there who successfully done this?

These are the steps i've done;

Change shell to bash and added /mnt/data as home directory

Code: Select all

chsh <username>
Changed:
Shell: /usr/local/bin/bash
Home directory: /mnt/data

Logged in as root, re-mounted disk as rw and installed perl.

Code: Select all

su
mount -o rw /dev/ufs/FreeNASs1a /
pkg_add -r perl

In Veeam shell under "Servers" i've added the server as a Linux server.
The SSH-connection works and it lists the files at the server.

Then under "Backup Repositories" i've added the server as a "Linux server".
The Location is set to /mnt/data and it populates both capacity and free space OK.

When finishing it states:
<Green check> Creating database racords for reository
<Green check> Backup repository has been added successfully

Next step is to create the backup job.
I choose a VM to backup.
Set backup repository to the repository i created.
Next, next, finish.

When i try to start the job i can see on the /tmp directory on my freenas box that it creates a file and runs a pl script:

Code: Select all

cd /tmp
ls
VeeamAgent91ddff83-ef1d-4643-880b-735bd0f8907b*
ps aux
perl veeam_soapafdb3eb7-b808-494c-9277-52af6471d79d.pl -d -c -l libafdb3eb7-b808-494c-9277-52af6471d79d -e /tmp/veeam_errorafdb3eb7-b808-494c-927

The job fails with:
"Failed to start client agent on the host "<ip>"
Timeout to start agent

Log shows:

Is there maybe some BSD-specific modifications to be done in veeam_soap.pl in veeam_soap.tar?
Or any other dependencies besides perl?

As a workaround i currently use freenas as a cifs-repository.

[flakrat]

I stumbled across this thread after upgrading from 5.x to 6.1 Veeam Backup and Replication.

I had to open tcp ports 2500 and 2501 to the Veeam server for backups to work. This wasn't the case with Veeam v5

The log file on the Linux server under /var/log/VeeamBackup provided the details.

# Veeam Backup Port (where x.x.x.x is the IP of the Veeam Server)
-A INPUT -s x.x.x.x -m state --state NEW -m tcp -p tcp --dport 2500:2501 -j ACCEPT

Hope this helps.

[yohay]

Struggled with this for a while to run Veeam backup to FreeBSD 11, 64-bit,

I've installing perl, sudo, bash and configured ssh to accept root password authentication things didn't run,

It seems like the FreeBSD in not executing the /tmp/veeamxxxx file as it is a 32-bit Linux file,

To solve this, Execute this in your FreeBSD shell and it all started to work:

kldload linux

Make sure you add to /etc/rc.conf

linux_enable="YES"

To make it run automatically with boot.

Hope it helps.