Backup WAN Target

[brecon]

Hi Guys

We are currently testing Veeam B+R to backup to a WAN target over private IP link ( which can range from 6Mb to 10Mb )

We have it working well at the moment to a CIFS share target after having created the initial full and manually moving it to WAN location and now increments are running over WAN

However we have hit the issue of the periodical synthetic full trying to process over the WAN

From reading the posts this is by design and the only solution in v5 is to switch to Reverse Incrementals or use a linux WAN target for backups as Veeam installs an agent to offload the processing of the synthetic backups to the target

I have tested the reverse incrementals and its just too slow and not feasible so we are going to test having a Linux WAN target and see what performance we get from that

Can someone confirm the following for me

- Will an ESX host as the WAN target be able to utilise the same Veeam "processing" agent that gets installed on the linux backup targets or do you need to use a "vanilla" linux target

- What is the best datastore to use on the linux backup target for performance - again from reading posts it seems like NFS datastore mounted on the linux backup target host ( or ESX target host if applicable ) is preferable

Any help would be appreciated

[Gostev]

Hi,

Yes/Yes.

Note that v6 will also support Windows-based agent-enabled backup repositories.

Thanks!

[Vitaliy S.]

Hello,

Another solution would be to place your backup server offsite, in this case all network traffic consumed while building synthetic full would be kept local.

Thank you.

[Gostev]

There are some major drawbacks of this other approach though (WAN traffic will not be compressed, possible vCenter/ESXi communication slowness/issues/timeouts, etc). So as a rule of thumb, I would definitely suggest always putting backup server close to the data being backed up.

[mschlott]

Can someone explain the concept of a backup target, and what processing agent brecon is talking about? I think I have been running my backups backwards. I have my Veeam server remote from my data center and run reverse incrementals to SAN storage. It sounds like it might be better if my Veeam server were in the data center and I had a NFS server at the remote site. I don't see how any sort of agent would come into play over a CIFS or NFS mount.

[Gostev]

We are talking about agent-enabled backup target above. With the current version (v5), it can only be a Linux server or "fat" ESX host (to which you mount NFS share). If you then pick that Linux host as the destination for your backups (having added the server to managed servers first), Veeam will deploy the small agent there in runtime that will be receiving the incremental data, and performing the full backup file processing locally (versus over WAN).

If you search this forum for offsite you should be able to find multiple existing discussions on this.

[mschlott]

Maybe I missunderstood. Would this be for replicas only and not backups?

[Gostev]

No, we are talking about backups specifically here.

[mschlott]

Thanks. I never noticed the dropdown on the backup destination page.

mike

[stevil]

Hi,

Yes/Yes.

Note that v6 will also support Windows-based agent-enabled backup repositories.

Thanks!

I'm glad you have kept to your word and included this in the new version.........excellent!

Even better will be the use of much smaller block sizes, I'm not sure this feature is in V6 or not

[Gostev]

Well, you can go down to 256KB block size from default 1024KB even today with v5, but not more than that and no changes in v6 here (but we do have plans).

[stevil]

Yeah, I was hoping to see something radical like 4k (as on some nSeries/netapp NAS's with WAN replication). Performance is an issue though I understand. I think our WAN links will be upgraded before we sort this issue out though