Veeam agent for Linux in cluster RHCS

[brodavdoug]

Hello all,

Even I know that the Veeam agent for Linux in cluster is not yet supported, I would like to know if someone already tried to use it in this mode ?

By my side I did it and I first tried to backup the active node (in "Entire machine" mode) and it works perfectly, so glad ! following this I tried to backup the passive node and the backup failed on the INACTIVE LMV2 volume as the very beginning.
The annoying point is : instead to skip the INACTIVE volume and continue to backup; the job stop immediately and nothing else is backuped.

All our last physical machine are under cluster and we would like to use only Veeam Backup for everything.

My servers are under redhat 6.5 with RHCS.
All feedback or advise will be appreciated until the so anticipated day where it will be supported come !

David

[PTide]

Hi and welcome to the community!

Have you tried to exclude the inactive volume from the backup job using "Volume backup" mode?

Thanks

[brodavdoug]

Not yet, but this is my next test, my only concern is: if I do this and my cluster ressources move to this node , this LVM will not be backuped anymore because it is exclude on this node..
Is it possible to change the policy with a pre-script option to remove or add this exclusion ?

Thanks again

[PTide]

Hi,

No, pre-job script won't work as new settings will be become effective only on the next job run.

Are both nodes identical? I mean is it really necessary to backup passive node? Because another workaround I can think of would be to install VAL on both nodes, configure pre-job script that checks if the node is Active. If the node is not active then the script just stops the job.

[brodavdoug]

Hello,

You are right about the need or not to backup the passive node but yes in our configuration we need to save it too. Sorry but what is VAL? Veeam Agent for Linux ? It's already install on both, yes.

The workaround I use now is to backup as "complete machine" the active node and use "Volume level Backup " for the passive node and exclude the INACTIVE VG.

I tried to identify where is the backup policy configuration file but without success (not even sure it works like that) it is probably in a Veeam database I guess.

I launch this:

Code: Select all

veeamconfig config export --file test

Like this I can see in xml format the policy details:

Code: Select all

<?xml version="1.0" encoding="UTF-8"?><Config Version="1" /><Repositories><Repository Name="Imported" Type="LocalDrive"><LocalRepoOptions Version="1" Location="/" DeviceMountPoint="" /></Repository></Repositories><Jobs><Job Name="Linux_Physical_SERVER01" creation_time="2017-03-17 15:00:42" RepoName="[SERVER00]Veeam_Data03"><JobOptions Compression="Lz4" BlockSize="KbBlockSize1024" PrefreezeCommand="" PostthawCommand="" PrejobCommand="" PostjobCommand="" MaxPoints="14" IsDedupEnabled="true" Schedule="true" RetryCount="3"><Schedule Hours="6" Minutes="20"><WeekMask Sunday="true" Monday="true" Tuesday="true" Wednesday="true" Thursday="true" Friday="true" Saturday="true" /></Schedule></JobOptions><ObjectsForBackup><Object ObjectType="SimpleDisk" RecordType="Include" Value="vg00" /><Object ObjectType="SimpleDisk" RecordType="Include" Value="vgDEV12U03" /></ObjectsForBackup></Job></Jobs><VbrServers><VbrServer Name="SERVER00" login="account_svc" domain="europe" fqdn="SSERVER00" password="" Endpoint="SERVER00:10002" /></VbrServers>

I could create two export files:
One with the exclusion of "vgDEV12U03"
One without exclusion of "vgDEV12U03"

The perfect solution would be to create a script running every 5 minutes for example on both node (I can do it easily in python for example) to detect if the VG ressource has been moved and automaticly change the policy configuration as command line and script the import of the xml file if needed

Thanks

[PTide]

The perfect solution would be to create a script running every 5 minutes for example on both node (I can do it easily in python for example) to detect if the VG ressource has been moved and automaticly change the policy configuration as command line and script the import of the xml file if needed

Yes, that may work, however keep in mind that if you import a config that contains VBR entry like this:

Code: Select all

<VbrServer Name="SERVER00" login="account_svc" domain="europe" fqdn="SSERVER00" password="" Endpoint="SERVER00:10002" /></VbrServers>

then you'll be prompted to enter the credentials, since we do not export passwords. There is a workaround - edit the config file so that it does not contain any VBR info between these tags:

Code: Select all

<VbrServers></VbrServers>

When done, adjust "objects to backup" part to your needs, and leave all other parts as they are. Now, if you import such config it will just update the existing job settings, without trying to update VBR serttings thus won't prompt you for credentials.

Thanks

[brodavdoug]

Hello,

Finally I found the good solution by create two jobs:

-First jobs is running a full backup every day automatically in “Volume Level backup” on both node with the clusterised volume unselected.
-Second policy with the logical volume clusterised only but starting manually.

Then I created a script with Pyhton to detect if the node is the owner of the clusterised volume
__________________

Code: Select all

#!/usr/bin/python
import os
import sys

f1 = os.popen (r"""lvscan |grep inactive""")

if len(f1.readlines())> 0:
        print "VG inactive on this host, not backup needed"
        sys.exit(0)

else:
        print "Start backup Job"
        os.popen("""veeamconfig job start --name Linux_physical _vgQV12U03""")
        sys.exit(0)

_______________

If yes the second policy is started by calling the second job name like this:
veeamconfig job start --name Linux_physical_vgQV12U03

Of course its not supported but it works like a charm..

Hope it helps.
David

[PTide]

Great job David! Did I get it right - you start this script as a post-job script is that correct? Also I can see two possible drawbacks of such design:

- it will work only with Server edition as multiple jobs are required

- since the shared volume is backed up by separate job there is no guarantee that inter-volume consistency will be kept (in case you have, say, a database that contains binary fields on one volume and an actual binary data on the VG)

Thanks

[brodavdoug]

Hello
Thanks
Yes, this second policy can be called by the post-job script of the first policy (full backup) or strarted after Rman backup or database export if need.
Well seen, as you said there is no consistenacy group between LUNs but my application support this; so it no matter for me but it could be a problem depending the applications I guess.

Now I will present Veeam backup to my manager and explain why and how we can replace our old existing backup application to Veeam backup

[PTide]

David,

Please contact support team directly to obtain a fix for the job failure with inactive LVM present issue and post your case ID.

Thank you

[brodavdoug]

Hello,

I opened a case on veeam support web site and I could get a fix for my problem.

Problem solved thanks again.

David