vROps Deployment model for One vCenter, two Datacenters, and three clusters

[MarcLaf]

I'm still unclear on deployment models to support this environment.

We have 1 cluster in one datacenter and 2 clusters in another datacenter all managed from 1 vCenter. I figured I need to deploy a main vROps node in one datacenter and a remote node in the other datacenter but how do I limit each instance to only monitor the cluster it's deployed on?

I tried deploying this before and after linking it to vCenter I think it was trying to gather stats from ALL clusters but it never worked properly so I wasn't sure.

Do I need to provide limited access for the account I use to link vROps to vCenter to see the chosen cluster? eg.

Account1 has access to cluster 1 in datacenter 1 and Account2 has access to cluster 2 and 3 in datacenter 2?

[daphnissov]

I figured I need to deploy a main vROps node in one datacenter and a remote node in the other datacenter but how do I limit each instance to only monitor the cluster it's deployed on?

Why would you want to do this? vROps is able to collect from all entities under your vCenter to which it connects. There's no need to segregate what it can collect to align with where the vROps nodes are placed.

What is the dispersion of your topology? How far are data centers A and B from each other?

[MarcLaf]

I thought performance would be better this way, that's why.

They are about 1000 km apart (600mi) and we have about a 20-30ms latency.

[daphnissov · Accepted solution]

Performance is really not going to matter because the data is serviced from the vROps node/cluster and not sourced on-demand from the monitored resources. At 600 miles and 20-30 ms of latency, you're probably fine not using a remote collector, but if you need to you can deploy a remote collector at the remote site. This still allows you to keep vROps at your "central" datacenter and send metrics back to it without as much chattiness.

[MarcLaf]

Thanks for the feedback. I think I'll deploy just a single instance at the main site to see how it fares. If I notice performance issues I can always scale out.