Question regarding proxies for backups to remote location

[Davd]

Hi,

We have all our VM's in Location A. In this location, we also have a VM with Veeam B&r v.6 with DAS storage used as backup repository and we run nightly backups of all VM's to this repository. We use 4 proxies for this and everything works fine. We are using Incremental backup mode, we have enabled synthetic fulls and we have enabled transform previous full backup chains into rollbacks

For security reasons, we are also required to run daily backups of all our VM's to a remote lacation (Loacation B). In this Location B we have the same setup as in Location A (a VM with Veeam B&r v.6 with DAS storage used as backup repository, 4 proxies and same backup jobs and backup settings). We have 1 gbps connection between Location A and B.

My question is: Where should the 4 proxies be located for backup jobs to remote Location B?

My initial thought was that they should be in Location A since transfer between a VM and proxie is the amount of data thats changed since last backup while transfer between proxy and backup repository is deduplicated and compressed which is great for remote backups. But once the backup job is completed and transform is executed, where does it get processed? Is it on the backup proxie server or Veeam B&R server? Does it matter for the speed of transform task if the processing server is in Location A or Location B?

[foggy]

My initial thought was that they should be in Location A since transfer between a VM and proxie is the amount of data thats changed since last backup while transfer between proxy and backup repository is deduplicated and compressed which is great for remote backups. But once the backup job is completed and transform is executed, where does it get processed? Is it on the backup proxie server or Veeam B&R server? Does it matter for the speed of transform task if the processing server is in Location A or Location B?

It's always better to have proxies closer to the source storage to benefit from the fast data retrieval via Direct SAN or Virtual Appliance mode. Since synthetic fulls and transformation activity take place locally on the backup target repository, it does not affect the location of the backup proxies.

Actually, the whole architecture, including the location of proxy servers, is covered in detail in the corresponding user guide chapter.

That said, I would say that your setup is a bit excessive. You do not actually need two backup servers to have two copies of backups (local and offsite). There are numerous existing discussions on how to perform offsite replication of backups, you can look through them to pick some ideas of what others do. Here is the couple of the most recent ones: Best backup setup for mirroring backup respository and v6 - How to have a local and off-site backup copy?

[rbrambley]

But once the backup job is completed and transform is executed, where does it get processed? Is it on the backup proxie server or Veeam B&R server? Does it matter for the speed of transform task if the processing server is in Location A or Location B?

I believe the transform and file management of the backup set is performed by the repository. Unless you have specified another system as a repository and are then using that repos in your configured job you are most likely therefore using the default repos - which is the Veeam server itself. We'll wait for clarification from Gostev on this to be sure.

So, as foggy mentioned, leave your proxies at Site A, but add a repository server in Site B. The backup target in Site B should be connected to the new repository and then make sure the site B repos is selected in the job settings. Now you have enabled compression across your WAN from proxies to repos, and you've enabled the file management locally to site B.

Of course the speed of the transform task is really impacted the most by the performance of your backup target. There is a lot of I/O involved in transforms. Search these forums for many threads already explaining this.

[rbrambley]

To be clear, even if you are using a CIFs share as a repos, this scenario will probably be better by putting another server in as a repos. Veeam can install the service on the linux or windows server, but it can't do that for a CIFs share. Configure the repos system to connect to the CIFs share. Actually, use NFS if you can because that is usually better.

This new repos could even be a VM - it's connecting to the actual backup target via networking (CIFs or NFS).

[Davd]

Thanks for your input. It confirms my thoughts about how things work. I will leave the proxies in site A and point to the remote repository. But a confirmation from Gostev about transform operation (on where its performed) would still be great.

We have already tried to synchronize backup files accross but without success due to the large data amount so thats the reason for why we perform two backup jobs. Creating a VM for backups and replicate it across is also a no go, and linux is not one of the systems we work with. So for now, we are ok with running two backup jobs. But if you guys come up with some other solution embedded in future Veeam versions that would also be nice.

[Vitaliy S.]

Transform is performed by a repository agent, no need to wait for confirmation from Gostev

[rbrambley]

Davd,

I just reread your original post trying to understand what your repos target is in SiteB.

So you will continue to use the DAS currently attached to the Veeam Server in SiteB as your repos? If so, are you are going to attach that storage to another system that is configured as a repos (and thus add the agent /service) with the Veeam Server in SiteA? That's an important design point!

[Davd]

Rich,

Let me try to clarify

We have the same setup in both Location A and B regarding hardware and configuration. But from actual Veeam VB&R application perspective, Veeam in Location A doesn't know existence of Veeam in Location B and vice versa. So all backup jobs that are performed on Veeam B&R server in Location A, only Location A proxies and Location A repository is used. And the same goes for Location B backups - backup jobs that are performed on Veeam B&R server in Location B, only Location B proxies and Location B repository is used.

So what I tried to find out is if Location B proxies are better off to be running on an ESXi in Location A due to the WAN link between the sites. And even if they should be running on an ESXi in Location A, those proxies would still continue to be tide-up for backup jobs performed by Veeam B&R in Location B.

The actual configuration of each locations DAS is following:

A DAS is connected to an ESXi server and on that server we have a VM with Veeam B&R (Win2008R2). DAS harddrives are used in one big LUN, then we have multiple 2TB VMDK's in that LUN and then all VMDK's are spanned to one big drive on the Veeam B&R VM. And this Veeam B&R server acts as Backup Repository.

I hope its a bit clearer now how our setup is.

[rbrambley]

OK. Understood. But, in your first post you said you are also required to backup VMs to siteB. Unless you actually meant "VMs at siteB" I understand this to mean you will have a job that backs up your VMs from SiteA to a repos in SiteB.

What I am trying to emphasize is that your current repository (DAS) in SiteB is configured to the Veeam server for SiteB today, right? Therefore that windows OS has the repos agent/ service running on it from the SiteB Veeam install so that means it can't be used as a repos for SiteA too. You will either need to uninstall Veeam from that server or dedicate another VM (or Box) to be the SiteA repos for these second jobs. Connect the DAS to this other system if you go with this.

Bottom line is that Veeam proxies and repositories are owned and used in jobs by the Veeam GUI / Console that the service was deployed from. They cannot be shared by multi Veeam installs.

OR

You could run everything from the SiteB Veeam server. To do that you would have to:

• unistall the SiteA Veeam services from the proxies and repos there
  add those same proxies and repos to the Veeam server in SiteB
  run local jobs (created on SiteB Veeam) in SiteA using only the SiteA proxies and repos
  run the A to B job (created on SiteB Veeam) from Site A proxies to SiteB Veeam server (repos already in place)

The gotcha with using SiteB for everything is that if the link goes down your SiteA local jobs won't run because the proxies and repos can't talk to Veeam SiteB (the brains!)

OR

Set up the SiteB Veeam server to have the A to B backupjob. That way you just need a proxy at siteA with the service from SiteB Veeam deployed. That may be an extra proxy from the ones you have today, or maybe you can uninstall siteA service from one already controlled?

IMO you are better off with 2 separate Veeam servers and working out the proxie / repos sets for the jobs between sites. Be sure to use the Enterprise Management server to work with and monitor the status of all jobs at both sites!

[Davd]

Thanks for you input Rich... I understand it now and have configured it according to your last statment with 2 separate Veeam servers and Enterprise Management server.

[rbrambley]

Awesome. Let us know how things are working!