[FEATURE Request] Adding multiple Proxies to AHV

[arogarth]

Please add a method to allow adding multiple Proxies to AHV Clusters! One of our Cluster is build of 19 Nodes and hosts over 600VMs!

To Process all over one Proxy - Specially if a new Setup will deploy with multiple VMs - The Proxy can become a bottleneck...

[HannesK]

Hello,
"can become" or actually is? How many CPUs & RAM did you give it? How fast is it processing data? I assume that you hit network limits at 10 / 20 / whatever Gbit/s?

There is a way in version 3 to deploy multiple AHV proxies with multiple VBR servers. There is still a 1:1 connection between AHV proxy an VBR in version 3. And one VM must only be processed by one VBR / AHV proxy.

Best regards,
Hannes

[arogarth]

That sounds good that there is a war do that.

I increased the vCPU and RAM to run 13 Jobs in parallel - but high number of vcpu is always a bad idea - multiple vms with less vcpu is better here...

[MNCowboy]

What timing! I came here to find out about this very thing.
We built a second Nutanix cluster. When adding to VBR, do I need a second proxy or will 1 AHVProxy suffice for both clusters?
We are moving VM's from Cluster1 to Cluster2, so there won't be any additional VM's.
TIA

[Mildur]

Each cluster needs a proxy.
If you have a new cluster, you need to deploy a new proxy and create new jobs.
If you move a vm to the new cluster, it needs to be added to a backup job on the new proxy. And a new Backup Chain will be created. No mapping is possible.

[MNCowboy]

Thank you Mildur!

[taurus1978]

i assume same rules apply as for vSphere Proxies. The Backup data moves from the source via the proxy to the target repository. As such only ONE Proxy for one single NUTANIX cluster seems to become a bottlenek in bigger environments. I think no customer wants a "monster Proxy VM" if environment scales f.e. to 1000 VMs and higher....
So a "scale-out" option for AHV Proxies would be fine.
Another consideration is: In an AHV environment VEEAM Proxy Appliance uses NUTANIX native Snapshots. Despite VMware, NUTANIX Snapshots are always "fulll lightweight" snapshots. How many load is put to the AHV proxy during backup. ? Similar load than in vSphere environments. ? And i assume VEEAM transfers the whole data of the snapshots through the proxy ?. If this is true scale-out options of the AHV proxy becomes even more important.

[HannesK]

Hello Patrick,
how many customers with more than 1000 Nutanix VMs do you have that face performance issues? While the feature request is valid of course, it's interesting for us to see how many customers really have issues with 1 GByte/s (expected speed for one proxy if the hardware is capable to deliver it) backup speed with one proxy.

While the option is not available, is the workaround I mentioned earlier possible in the meantime?

I'm not sure I understand the second question. While the snapshot technology is different than to VMware, we of only read changed blocks instead of the whole data. Change block tracking is essential an we only do support hypervisors that have the CBT feature natively available (except legacy Hyper-V where we wrote the CBT driver on our own). The way we do it is the way to leverage CBT, so we cannot do it like on VMware.

Best regards,
Hannes

[taurus1978]

Hello Hannes,
i reference to the proxy "limitation" for AHV clusters in terms of "scale-out". Like we can do it in VMware Environments. -> "proxy per host" approach. Currently if we need to scale the AHV proxy we need to do a "scale-up" instead of a "scale-out". This may result in a large VM for the AHV proxy. So at some point we have a limitation regarding parallell taks because we dont want to further scale up the CPU or memory because we dont want a "monster VM" .

Additinonally some clearance to the folowing topics would be good:
- How exactly does the AHV VEEAM proxy transport data to the target repository. ? -> I assume the VEEAM data mover service is doing the job. And talking to the target data mover. How much load or data is put to the proxy per task ?

- How much load is expected to the AHV proxy when using compression/de-dup/ or encryption.? -> i assume similar rules apply than to vSphere/Hyper-V or physical proxies.?

Finally i think some guidelines in the official docs would be good.

[HannesK]

Hello,
I think we agree that this is a limitation. I just liked to know, how often you run into technical issues by doing "scale up". "don't want" is more a "personal preference" instead of "technical issue" from my point of view

Traffic flow is described in the user guide

Yes, load expectations are the same like for all other image based backups.

Best regards,
Hannes

PS: please use separate topics for questions that are unrelated to the thread (I will answer them in other topics, but not here)

[eduzimrs]

This feature would be very in handy, we have customers with cluster containing more than 800 VMs and 300TB, the processing rate with only one Proxy is making things hard and missing RPOs.