Feature request: Small VBK file reader with restore option?

[gollem]

Hi,

We've got 21 sites across Europe, they are all identical in its setup:
- On-site Primary server: This server hosts all live servers and a veeam backup proxy
- On-site Failover server: Everything which is running on the Primary server is replicated every hour to this server
- On-site NAS: has backups for daily and weekly jobs and hosts replication data for the site
- Remote (Datacenter): Daily Replication of all servers on the Primary server to the remote location
Locations are connected to the Datacenter with 4,8 or 10Mbit links

The setup is quite beneficial for us as we've got a hardware independent setup without a single point of failure (its a manual switch over, but that is about it), only in case of a complete site loss we rely on the replicated servers in the DC

Centrally in the DC we've got a Veeam server which talks to the proxies and initiate commands.

This method works quite well and has proven to be stable and reliable

However when restoring files it takes quite a bit of time for the restore console to pop-up (several minutes), I assume the Veeam server in the DC needs to read the windows file system from the backup and has to transfer it across the WAN before its shown. Restoring a large portion of files is also slow as data is traveling across the WAN

Assuming I want to restore a folder with 5GB of data from a file server: the files are being read via the location proxy, then send to the DC and then send back to the file server on site, resulting in 10GB which has to travel across the WAN, with slow WAN links this is a time consuming restore.

Alternative option 1: I can deploy a linux file recovery appliance and then pull the file locally using FTP, this works but is far from ideal.
Alternative option 2: I could install a free version of Veeam B&R on each proxy and import the backup files, this also works but again far from ideal when updating the proxies

Ideally there would be a small application (portable, no installation required) which can read the backup files VBK and could restore those.

Is this something which can be looked in to please?

Thanks!

[PTide]

Hi,

I'd like to request some additional info if you don't mind. Where do your mount servers reside?

Thanks

[gollem]

Many thanks for your response!

Denmark: Location which I'm currently testing with
Datacenter: (AT&T Datacenter in the United Kingdom)
The Denmark site has got a 10up/10down mbit link and the ping delay to the datacenter is 43ms

I've setup the backup job as the following:
(Storage section)
Backup proxy: 1 proxy selected, Denmark server
Backup repository: Denmark repository selected

And for the repository, its a Shared folder which resides on a Netgear NAS onsite (SMB)
Gateway server: Denmark proxy server
Mount server: Denmark proxy server


When restoring I've had a look at the log, and although the mount server should have been the denmark proxy, its actually selecting the Datacenter server.

When looking at the performance monitor on the datacenter server, I can see that veeamagent.exe communicates with proxy server on site.

Could the problem be due to that we are using a NAS (shared folders) for our backup file storage?

FYI: Just loading the recovery window to the point where I can select the file to restore took just under 15 minutes to load

[PTide]

Ok, now I see.

Assuming I want to restore a folder with 5GB of data from a file server: the files are being read via the location proxy, then send to the DC and then send back to the file server on site, resulting in 10GB which has to travel across the WAN, with slow WAN links this is a time consuming restore.

Almost correct In fact backup files are mounted on the machine where FLR was started, however not all backup data is transferred but only data that is required for guest filesystem browsing. Once you select files for restore the second mount point is created on the mount server that is assigned to the backup repository where the backup belongs to.

There is a workaround that should make it possible to browse for files in the backups without mounting them on VBR:

1. Enable Guest OS indexing in the backup jobs settings.
2. Use EM manager for File-level restores.

Thanks

[gollem]

Well, since the loading of the backups was so slow I just assumed restoring would be causing the same bandwidth troubles, so i never went down that path and hopped straight on the restore appliance option ... sorry about that

There is a workaround that should make it possible to browse for files in the backups without mounting them on VBR:

1. Enable Guest OS indexing in the backup jobs settings.
2. Use EM manager for File-level restores.

Both work arounds where not really an option, but your previous post did get me thinking.

I was running the console on the server in the Datacenter so restore paths where like this:
- Launched restore, backup files are mounted to the DC Server, (resulted in 15 minutes waiting time before I could select files)
- Restored files which was done by the Mount server (which also is the proxy) and acceptable speeds

So now I've done a test where I installed the console on a computer in the location and this indeed reduced the initial loading times drastically. To get to the file select view is just a minute or so.

I always assumed that a full install would be required to work with the backup files, so now I just installed the console on the proxy which is a good solution for us. Maybe there can be an option when deploying the proxy that the console is also installed (with a optional tick-box, check to also deploy the console) ?

[PTide]

Both work arounds where not really an option, but your previous post did get me thinking.

That were just two steps of the same workaround actually Why is EM not an option in your case? It was designed to simplify FLR procedure for indexed backups.

I'm afraid that it won't be just a checkbox since VBR console requires a lot of additional stuff to be installed, which means that either the console installation package should be already available on the proxy, or all components have to be pulled from somewhere. In both cases additional actions from an operator are likely to be needed. Given the above, I don't understand how such a tick-box would be easier than manual console deployment. Anyway your request has been noted, thanks for sharing!