Veeam and vCenter backups

[emtunc]

Before I explain the issue I observe in our environment it is best to briefly detail the set-up:
i. We have a number of VMs sitting on an ESXi 5 host
ii. I have Veeam B&R and vCenter sitting on the same VM (we shall call this Server A)

The issue I am observing is that the incrementals for Server A are huge. Even after performing an incremental an hour or few later the incremental shoots to tens of GB's.
I have left Application Aware Processing enabled because I read a few posts here saying that it should be working fine with the latest B&R suite.

It it worth be turning AAP off? What effect would this have if I needed to restore this VM in the future?

The back-up data store for this particular job sits on an encrypted USB drive. Also other jobs do not experience this behaviour so I don't think it is to do with anti-virus or similar product scanning files and updating time-stamps.

Ideas? I'm free to suggestions as I am still trialing the product.

[Vitaliy S.]

Hi Mikail,

Application-aware image processing doesn't have any effect on the number of changed blocks on your VM. If you decide to turn this option off for highly-transactional applications (such as SQL Server), then it might lead to situation when you will have inconsistent backups of your applications.

Also other jobs do not experience this behaviour so I don't think it is to do with anti-virus or similar product scanning files and updating time-stamps.

Do you backup any other VM with SQL Server installed on it? Times-stamps, defragmentation, virus scans are not the only reasons for having large increments, SQL Server as well as Exchange VMs are "known" as applications that relocate/change virtual disks blocks quite intensively.

Thanks!

[AndyGDIT]

Are you using the hotadd feature? Are you running more than one job at a time?

I have seen where one of our field sites was backing up the veeam server while it was also backing up other servers. This lead to that backup being huge (because it had the other disks mapped to it). I have since told them to stop backing up the veeam servers as they are just proxies, and they are easy to replace incase of a failure.

Is it possible for you to have a seperate server for your Veeam B&R

[emtunc]

Hi,
Thanks for the replies. I will provide more information just to make sure I am not confusing my self.
Below we have the results of a successful incremental run. The second row is the Veeam+vCenter VM. As you can see the disk capacity of the VM is 200GB of which Veeam is reading 195GB which is causing it to take several hours... whether the 'Read' column means the number of changed blocks the OS is reporting I don't know.

Compare this with the last row (CRM server) which is reading only 20GB (20GB of changed blocks on the server?) and transferring 12.6GB of data.
So why is it that the Veeam+vCenter server reads so much but transfers little?


Size Read Transferred
200.0 GB 3.7 GB 1.7 GB
200.0 GB 195.1 GB 1.6 GB
50.0 GB 2.7 GB 1.2 GB
120.0 GB 20.0 GB 12.6 GB

@Andy, what is this hotadd feature? I could not find a setting for this anywhere.
This is the only job running at the time. There is a possibility to move the Veeam server to a dedicated VM but don't want to do that and find the issue still exists.

[veremin]

If you backup your VB&R server using default proxy server which is deployed on it, then CBT is automatically disabled and it’s VB&R proprietary mechanism that is being used. If CBT doesn't work for some reason, VB&R scans through the VM image and a checksum (hash) is calculated for every data block, then it verifies what blocks have been changed on your VM compared to the VM image stored in the backup file. I believe that is a reason why you see such high read value.

Thus, it's worth using another proxy server(probably, in Hot Add mode) to back this VM up.

Hot-Add (Virtual Appliance) is a proxy mode which is recommended and can only be used if the backup proxy is deployed on a VM. It uses the SCSI hot-add capability of ESX hosts to attach disks of the backed up VM to the backup proxy VM. In this mode, VM data is retrieved directly from storage through the ESX I/O stack, instead of going through the network stack, which improves performance. More information regarding it can be found in the corresponding User Guide and sticky FAQ.

Additionally, please be aware that with the introduction of configuration backup feature which is responsible for retrieving configuration data from the SQL database there is no longer need for backing up VB&R server as a whole.

Hope this helps.
Thanks.

[emtunc]

Thanks Vladimir. I think you are correct about the CBT being disabled. It also explains why so much data is being read as compared to the other VMs.

I believe all VMs except the Linux VMs are using hot-add (checked logs).

if the Veeam server is not backed up, what would be the fastest way to recover VMs in a disaster scenario assuming the Veeam server also dies?
I briefly read of the extract.exe utility but not sure if this would be the fastest route in recovery mode. Perhaps re-installing B&R and recovering from the UI? Would the config database require a restore for us to do this?

Also, do you see any technical disadvantage of installing B&R on a non-server OS? I can't think of any but perhaps Windows Server has more API's available than Windows 7 for example which may affect B&R. Just putting it out there.

[veremin]

With configuration backup if need be you can just re-install VB&R from the scratch and then import configuration. That’s all.
For more information regarding this functionality please see corresponding User Guide (p.305).

As to your second question, you can install VB&R on any OS that is mentioned in list of supported OS (p. 63) without any issues.

Hope this helps.
Thanks.

[AndyGDIT]

Hi Mikail

What Vladimir said is true. The job configuration is set up on the SQL database so there is no reason to backup the VBR server, I also have my SQL database on a seperate server, in which I do backup

What I have is a DataDomain DD890 in which I use two Linux servers as NFS proxies to the Data Domain. I then setup the job to point the job to that specific spot on the NFS server. The backup files are then replicated over to a DR site.

I have a "master" veeam server which has all of the jobs on it and each of the jobs have certain proxies that can run the backup. (I have several clusters which the ESX hosts have different disks to them). I have a total of 8 proxies that backup over 300 VMs.

If I ever have a Disaster (God Forbid) All I would need to do is spin up a VM, install the Veeam B&R and remap to those NFS proxies and re-import the backups for the configuration DB restore that any other servers that are critical, then I can re-create the jobs

I have never installed Veeam on a non server OS but I have read that some users are doing it.

[Vitaliy S.]

Perhaps re-installing B&R and recovering from the UI? Would the config database require a restore for us to do this?

Configuration import is not required. If you want to restore VMs you've been backing up with Veeam, just import our backup files to the backup console and start the restore operation.