Repository and proxy design questions

[ChrAlex]

Hi,
I have a two-site setup with a primary site and a DR site. Primary site consists of 3 ESXi5 hosts and EqualLogic iSCSI storage, DR site consists of 1 ESXi5 host and EqualLogic iSCSI storage. Everything is in the same vCenter. 1Gb dedicate fibre link between the sites for isolated iSCSI-traffic.
On the primary site I run B&R in a vm. The backup storage is connected in the vm using Windows' iSCSI initiator, meaning I have E:\<backup folders>, where E: is located on EqualLogic on the DR site.
What do I need to do when deploying additional backup proxies in the primary site, since they won't have direct access to the repositories? This being because the repositories are seen as local drives/folders on the backup server.
Do I need to add them as CIFS/SMB repositories to give a new vm acting as a backup proxy access?

Regards,
ChrAlex

[Vitaliy S.]

Hi Alex,

Have you considered using dedicated VM acting as a backup repository which in turn will be connected through iSCSI initiator to the DR site? In this case you can deploy as many backup proxies as you need and point them to this backup repository VM.

Thank you.

[ChrAlex]

Thanks for the suggestion, Vitaliy! That might be easiest thing to do.
The reason for the current design is because B&R was upgraded from v5 on this server, it therefore holds the job configuration and everything else.
If a added a new VM at the production site as a backup proxy with the current setup, would it work? I don't see how the new proxy would connect to the backup repositories when they are local attached drives in the VM where B&R is installed.

Thanks!
Regards,
ChrAlex

[foggy]

Alex, in the current setup, the role of the dedicated VM Vitaliy is talking about is implemented by the backup server itself. Just specify the Default Backup Repository as target for the new proxy, that should work. Thanks.

[ChrAlex]

Thanks foggy! That's definetly the easiest solution, so I'll give that a try. I still don't understand how the additional proxies will access the repositories though. Are there any drawbacks creating repositories as CIFS/SMB shares instead of "direct attached" drives?

Thanks for the quick replies so far!

[Vitaliy S.]

Just to be clear...

Why do you refer to multiple repositories? If you dedicate 1 VM to act as a repository, then point additional proxies to use this repository (given that you have a network connection from backup proxies to this repository VM), there will be no issues.

All backup proxies will treat your "iSCSI mapped drive" as local drive on repository server. Hope this helps!

[ChrAlex]

Before the upgrade from v5, each backupjob used its own folder, eg e:\<name of job 1>. After the upgrade this resulted in several repositories, one for each folder. I can probably make this easier by making the root of the drive a repository and let B&R handle the making of directories for each backup job, resulting in one repository instead of several. But for the functionality you describe I guess this doesn't make any difference?
Thanks!

[Vitaliy S.]

Yes, you can either consolidate all repositories into one, or leave it as it is now.

[mclemen2]

[merged]

I've read through the FAQs and something still doesn't seem clear to me.

We have 2 proxies that are connected to the same FiberChannel SAN to use as a backup target. The same volumes on the SAN are presented to both proxies. I can attach/mount a volume to one server, and create a repository on that volume. I can create a backup job and point it to the repository. However, what happens when the proxy server that does not hold the repository gets assigned the job? Does the proxy performing the backup send the data over the network to the proxy holding the volume? Or does the multiple proxy scenario not work in that instance? Would I have to set up the repository as a CIFS share to enable both the "host proxy", and the "remote proxy" to access the repository.

For example:

proxy1 has LUN0 attached as the G: drive. I create a repository REP0 that points to the G: drive. I create a backup job and point it to the REP0 repository and enable both proxy1 and proxy2.
If proxy2 is too busy to be selected (as the host of the repository, it should be in the first choice if I read how the different proxies are ranked correctly) and proxy1 is selected, how does the data get sent to the G: drive on proxy1? Is that not possible without creating a CIFS share for G: and setting the repository to point to the CIFS share? In which case, what's the point really if proxy1 is already so overloaded that another proxy was assigned the job.

[dellock6]

Hi Michael,
I do not think is possible to share the same LUN to be configured as local disk for two different repositories. Repositories needs to have their disk available when performing backups, what you are describing is a sort of clustering between repositories, and is not supported.
A better design, imho, would be to remove repository roles from proxies and add a designated server to act as repository, or in your design assign different LUNs to the two proxies/repositories.

Luca.