VxRail/vSAN backup with Veeam

[StanoSedliak]

Hi,

do you have maybe some experience with backup of VxRail/vSAN, about some best practice or tips, experience about the speed how its going.
Setup:
Physical backup server with intern/local disk as backup storage
On every VxRail node a VM as veeam proxy (5 nodes together) with Virtual Appliance Mode.

Thank you!

[HannesK]

Hello,

Physical backup server with intern/local disk as backup storage

all-in-one box, that's what I would also go for

On every VxRail node a VM as veeam proxy (5 nodes together) with Virtual Appliance Mode.

if you like. I would do it without and see whether it's fast enough. It's a personal preference, but I don't like HotAdd, because of the hot-add overhead, which can take longer than the backup itself.

Best regards,
Hannes

[Giacomo_N]

Hi, I've a similar setup since January 2021:
VxRail stretched cluster 3+3 nodes and two external R740 with vmware and Win2019VM as backup server/repository/tape, another Dell in another site (40Gbs connected) only to copy jobs.
The best practice is 1 proxy per host and Virtual appliace mode, I tried hod add, but is slower in my enviroment.
I've set 6 concurrent tasks per proxy, more tasks not affect the backup time, but depends the hardware, you must try.

[HannesK]

The best practice is 1 proxy per host and Virtual appliace mode,

may I ask where you found that? I tried to remove / change all documents that said that, because we dropped that recommendation some years ago.

[Giacomo_N]

Hi HannesK, really no official document because, tell me if I'm wrong, Veeam didn't write anything about Vxrail/Vsan.
I inherited the best practice write from and with Nutanix (used in our enviroment from 2016 to 2021) by Luca dell'Oca around the concept of HCI/data locality.
The concept was that the proxy process the data of the VM's stored in the same specific host of the proxy, this is also possible making a "should" rule in the Vpshere cluster to keep, as possible, every proxy to the specific host.

Tell me/us if I'm wrong and if there are new directions about VxRail/Vsan.

Thank you!

[Giacomo_N]

There is also this old post, answered by Luca : vmware-vsphere-f24/best-practices-for-v ... 38854.html

[StanoSedliak]

may I ask where you found that? I tried to remove / change all documents that said that, because we dropped that recommendation some years ago.

If you mean the info about 1proxy=node its here:

https://helpcenter.veeam.com/docs/backu ... ml?ver=110
Virtual Appliance Mode for VMs on VSAN
The backup proxy VM must reside on any of ESXi hosts connected to a VSAN cluster.

Hi, I've a similar setup since January 2021:
VxRail stretched cluster 3+3 nodes and two external R740 with vmware and Win2019VM as backup server/repository/tape, another Dell in another site (40Gbs connected) only to copy jobs.
The best practice is 1 proxy per host and Virtual appliace mode, I tried hod add, but is slower in my enviroment.
I've set 6 concurrent tasks per proxy, more tasks not affect the backup time, but depends the hardware, you must try.

Thank you for info, did you try to run the backup without proxy on every node? If it was big difference.

[Giacomo_N]

@Stano
Good to know that the direction is present in the current documentation.

Thank you for info, did you try to run the backup without proxy on every node? If it was big difference.

Really no, after coming from Nutanix I keep the same "logic" about proxy numbers, I played only on how many cores and tasks to use in the six proxies.
I'll try to deploy or disable some proxies in the next days, but actually I've 13-14 minutes of daily/incremental backup and 2 hours of full backup during w.end keeping the infrastructure reliable, in my opinion is fast!

[StanoSedliak]

@Giacomo_N - uf it's really quick, can tell me please the Processing rate what you get with your setup please?
And how much VMs are in rou backup please?
Is there any reason why you have the backup server on your R740 as VM and not directly OS installed?

Thank you!

[Origin 2000]

Deployed a VxRail 3+3 stretched Cluster some time ago as well es several standard vSAN Clusters. No problem at all when it comes to Veeam. We use HotAdd and implement a propper monitoring when it comes to unwanted or hidden Snaps. In most case we have less Veeam proxies as Hosts but it depends on how many phys. repository server a customer have.

Regards,
Joerg

[HannesK]

yes, the explanation in the documentation is correct. The point is, that this saving of bandwidth became irrelevant with increased bandwidth over the last years.

I will ask the documentation team to write it clearer. "Any" means, that for HotAdd, the proxy needs to be located on a host, that has access to the VSAN datastore. It does not mean "every host".

[Origin 2000]

Is there any reason why you have the backup server on your R740 as VM and not directly OS installed?

There is no right or wrong there. If you have VBR+EM in a VM it share the same benefits as every other VM
- You can create a snapshot before upgrading/patching OS or VBR
- You can assign ressources when neeed to the VM
- Not bind to the HW warranty of the server
- Is protected by vSphere HA

Regards,
Joerg

[Giacomo_N]

@Giacomo_N - uf it's really quick, can tell me please the Processing rate what you get with your setup please?
And how much VMs are in rou backup please?
Is there any reason why you have the backup server on your R740 as VM and not directly OS installed?

Thank you!

Hi, we've two simultaneus jobs of 41+2VMs(I need different retention) with 3.7TB+2.6TB processed running at 8PM, the last Friday transfer was 765MB/s+570MB/s taking 14 minutes.
I've a third single Job of 10 VMs running at 5AM with 2TB processed, the last transfer speed was 1.5GB/s and taking 13 minutes.

The reasons to keep the Veeam virtual are, as write before @ Origin 2000, the snapshots (I can't take hot snapshot due to Tape/SAS passthrough), I can keep the VM off during the day and I have other 2 VMs in this mini external cluster like monitoring tool and log server.

[Giacomo_N]

Deployed a VxRail 3+3 stretched Cluster some time ago as well es several standard vSAN Clusters. No problem at all when it comes to Veeam. We use HotAdd and implement a propper monitoring when it comes to unwanted or hidden Snaps. In most case we have less Veeam proxies as Hosts but it depends on how many phys. repository server a customer have.

Regards,
Joerg

Hi, do you have a customer with 3+3 stretched and perform better with hotadd?
For sure depends also by the network speed, in my enviroments I've 10Gb/s, may be @25 or 40 hotadd perform better.
Another aspect is the size of the processed VMs, we've 50 VMs "small" and hotadd taking minutes to start..

[StanoSedliak]

@Giacomo_N - thank you for detailed info about speed, the proxies you are using are linux or win?
As you wrote hotadd was slower for you, are you now running it in NBD/network transfer mode?

[Giacomo_N]

You're welcome

I've a Win proxies.
With hotadd I mean Network/NBD, in my enviroment is slower compared to Virtual appliance mode.
Probably, I'm wrong with the terminlogy...