Veeam Proxy Selection

[Novox]

I am currently running VBR in a windows server VM.

I have an NFS mounted NAS for a backup repository, and a Tape Server on another physical windows server.

Since I have NO proxies installed (except for the VBR server), my backup-to-tape job seems to be assembling synthetic fulls on the VBR server (I see read/write bandwidth). I assume VBR-proxy is reading VBKs/VIBs from my NFS mounted NAS to assemble the synthetic full and writing the generated synthetic full VBK to the physical tape server.

If I make the physical tape server a proxy, and the backup-to-tape were to use it, I assume I would get better performance, since the bits aren't traversing the network twice (read then write).

My question is... how do I know the job will run on the physical tape server/proxy? I found the following explanation from here (vmware-vsphere-f24/vmware-frequently-as ... tml#p39952) and other posts:

Q: How is the backup proxy selected in cases when there are multiple proxies available?
A: For each VM in the job, backup server classifies all backup proxies available to the job into 3 groups by access type they have to the VM's disks: SAN (1), hot add (2) and network (3). Proxy with least tasks currently assigned is picked from group 1, and if responding, VM processing task is assigned to this proxy. Otherwise, the next least busy proxy is selected from group 1. If all available proxies group 1 are already running max number of concurrent tasks, selection process switches to group 2 and repeats. Group 3 is only used if groups 1 and 2 have no responding proxies. For group 3, proxy subnet is also considered when picking the best proxy.

My current VBR proxies with the NBD/Network transport. I assume the new VBR would also use NBD/Network. Since the subnets are the same, it would seem which proxy that would get used is indeterminate (if neither are in use at the start of the job). How do I steer this back-to-tape job to the physical tape server proxy?

The network subnetting does seem to allow me to define a single host subnet (/32). Would setting the physical tape server proxy's subnet to be the IP of the physical tape server itself accomplish this? If not, are there any suggestions? I don't want the physical tape server proxy to sometimes be used for this job, I'd want it to always be used for this job (if my other assumptions are correct).

Thank you!

[HannesK]

Hello,

I assume VBR-proxy is reading VBKs/VIBs from my NFS

the repository / gateway server is doing that. I assume that your gateway server is set to automatic. https://helpcenter.veeam.com/docs/backu ... er=100#nfs

I recommend this section of the FAQ: http://forums.veeam.com/veeam-backup-re ... tml#p94869

To improve overall speed I suggest three things:
1) make the tape server also a proxy server: https://helpcenter.veeam.com/docs/backu ... ml?ver=100
2) switch your backup mode to "active full" instead of "synthetic full" to avoid slow synthetic processing: https://helpcenter.veeam.com/docs/backu ... ml?ver=100
3) set per-VM backup file for the NFS repository if not already done (for better parallel processing with the two proxies): https://helpcenter.veeam.com/docs/backu ... ml?ver=100
3) keep all other settings default (that includes automatic selection for the gateway)

Best regards,
Hannes

[Novox]

Thank you, Hannes.

Re: #1 below, I will make my tape server a proxy server, but given this is the proxy selection process:

For each VM in the job, backup server classifies all backup proxies available to the job into 3 groups by access type they have to the VM's disks: SAN (1), hot add (2) and network (3). Proxy with least tasks currently assigned is picked from group 1, and if responding, VM processing task is assigned to this proxy. Otherwise, the next least busy proxy is selected from group 1. If all available proxies group 1 are already running max number of concurrent tasks, selection process switches to group 2 and repeats. Group 3 is only used if groups 1 and 2 have no responding proxies. For group 3, proxy subnet is also considered when picking the best proxy.

When and how does VBR know to use the proxy on the tape server for tape server jobs? There must be additional proxy selection criteria, no?

[HannesK]

Hi,
automatic selection is default in the job settings. Decision about proxies is made during job run: https://helpcenter.veeam.com/docs/backu ... ml?ver=100

That's why I asked to keep everything else on default

Best regards,
Hannes

[Novox]

Hannes, how can I tell which proxy a backup-to-tape job ran on?

Backup-to-tape jobs don't have "Using backup proxy ... for disk ... [nbd]" in their job history, so how can I ensure my backup-to-tape and synthetic full creation are happening on my physical backup server (where I installed a second proxy on your recommendation).

Thank you!

[HannesK]

Hello,
no proxy runs backup-to-tape jobs. The tape server is the only role that can write to tape (from the repository or gateway server in your case) https://helpcenter.veeam.com/docs/backu ... ml?ver=100

Synthetic fulls are created by the gateway server (which I recommended to keep "automatic") https://helpcenter.veeam.com/docs/backu ... ml?ver=100 - there is a synthetic operations section on that page. And I recommended active full to improve performance (assuming that your production environment is fast enough)

Best regards,
Hannes

[Novox]

Thank you, Hannes.

It sounds like I need to re-architect my backup environment a little after introducing tape backup...

Currently, I am storing as much as I can on spinning drives so I have "forever forward incremental" with 60 retention copies. Periodic "Active Fulls" would convert my chains into "[NON-forever] forward incremental" and use up much more space.

However, if tape is now my long term storage, based on what you've saying, I should switch to "forward incremental" with periodic Active Fulls:

• Periodic Active Full prevents the need for synthetics for tapes

• Periodic Active Full prevents the need for defragmentation jobs

• Periodic Active Full removes all backup chain transformations (no more "merging" and a more predictable job schedule)

• Periodic Active Full may make my backups more reliable in that with one VBK and 60 VIBs, one assumes there is a greater chance for a restore operation to fail, i.e. if any interim VIB is corrupt? - check me on this please

• Periodic Active Fully may remove the necessity for "Storage-level corruption guard"? (given one assumes VBR will remedy Active Full creation issues on-the-fly)? - check me on this please

This is all at the expense of extra storage on my spinning drive media (now compensated by my tape infrastructure).

Does this sound like a solid plan?

Thank you!

[HannesK]

Hello,

to assemble the synthetic full

that made me think, that you already use synthetic full. As you use NFS, active full would not cost any disk space in that setup. That's why I recommended it.

It's just a small wording difference that created the misunderstanding. https://helpcenter.veeam.com/docs/backu ... ml?ver=100 is the real question

For synthetic operations, you can configure the mount server to your tape server: https://helpcenter.veeam.com/docs/backu ... ml?ver=100

What you write about your plan: everything correct, except the Storage level corruption guard is up to "personal preference". If the backup storage is broken, then it can still detect issues even with active full backups.

Best regards,
Hannes