VMware Cloud Community
halbewurst
Contributor
Contributor

Right Sizing of vRops

Hi all,

i am right in front of a (for me tricky) situation. The company i am working for has a worldwide deployment of vCops servers (Analytics, UI plus Database). As we have 6 Datacenters (10 Hosts each Datacenter in 1 Cluster with around 1000 VMs) around the world we have a independet vCops running on each of this Datacenters. For a combined reporting this is not the best as you can imagine. The plan now is to move to vRops picking one of the Datacenters to be the centralized Datacenter for vRops (Nodes) and all other Datacenters will send their Data (through remote collectors) to the centralized Datacenter. I am right in the middle of reading and thinking and as a first step i checked how many objects we have in total and how many metrics are collected in total. Objects are 55000, Metrics are 1312500 (numbers are from vcops audit report).

With that numbers i started checking in the sizing guide from vmware (VMware KB: vRealize Operations Manager 6.1 and 6.2 Sizing Guidelines).

I am now struggleing with the sizing guide, especially with "Multi-Node Maximum Objects per Node" which says a max of 5000 for medium and 10000 for large.....Do i understand it correct when saying, having 55000 Objects worldwide (growing) i need a minimum for 6 large Nodes to cover all the Objects?? Metrics should not be a problem but i do really struggle in understanding the Objects piece.

About the main setup of our vmware environment i do not have control as this is handled by other teams, also i have currently no idead if our network will be able to manage the impact of the data which would be send around by the collectors.

If anyone has any other advise or experience with such deployments, any help is welcome. The main goal is to have one single point to login and to maintain.

Thanks for your help in advance!

Reply
0 Kudos
4 Replies
sxnxr
Commander
Commander

First of i am not sure how vrops will handle the time stamps from the remote collectors that are in different time zones or if it is even supported. I know that the data, master and replica must be in the same time zone.

As for sizing i have a 6 node cluster (1 x Master, 1 x Replica 2 x data with 2 x remote collectors) this is collecting from 5300 x VMs and only using the vcenter adapter. I also have the EPO agent deployed to 4 VCSA servers. the master, replica and both data nodes are configured with 4 vcpu and 48 GB ram and i have no performance problems.

Reply
0 Kudos
halbewurst
Contributor
Contributor

Sxnxr,

thanks for brining up the timestamp thing...totally forgot about that. One more point on the agenda 😞

Are all the Nodes of your Cluster at the same Datacenter or do you collect with the Remote collector from another Datacenter if i understood correct the Remote Collector can also be used to be able to collect more metrics at the same datacenter.  BTW how many Metrics do you collect with that setup?

Reply
0 Kudos
sxnxr
Commander
Commander

We have 2 datacenters in different time zones so i have a 6 node (4 nodes and 2 remote collectors) in each datacenter. The 4 Nodes cluster is for the main environments and i use remote collectors for different environments that have firewall between them and the main network (using RC's only requires 2 ports open in the firewall)

This is form the audit report on 1 datacenter

Resources Collecting 7584

vCenter Adapter 7813

 Network Folder 4

 Virtual Machine Folder 141

 vCenter Server 5

 Datacenter 5

 Custom Datacenter 5

 Virtual Machine 5295

 VM Entity Status 16

 vSphere Distributed Port Group 140

 vSphere World 1

 Datastore 1639

 Host System 458

 Host Folder 16

 Resource Pool 13

 Cluster Compute Resource 55

 vSphere Distributed Switch 20

 Metrics

 Metrics Configured 15129923

 Metrics Collecting 2547071

 Super Metrics 60443

 vCenter Operations Generated 75870

Applications Count 0

Custom Groups 44

Resources Collecting 7584
vCenter Adapter 7813
 Network Folder 4
 Virtual Machine Folder 141
 vCenter Server 5
 Datacenter 5
 Custom Datacenter 5
 Virtual Machine 5295
 VM Entity Status 16
 vSphere Distributed Port Group 140
 vSphere World 1
 Datastore 1639
 Host System 458
 Host Folder 16
 Resource Pool 13
 Cluster Compute Resource 55
 vSphere Distributed Switch 20
 Metrics
 Metrics Configured 15129923
 Metrics Collecting 2547071
 Super Metrics 60443
 vCenter Operations Generated 75870
Applications Count
0
Custom Groups 44
Reply
0 Kudos
mark_j
Virtuoso
Virtuoso

Regarding the timezone, set all of your vR Ops appliances to use GMT. This way all the logs are in the same timezone and it'll make your life far easier.

On consolidating your vR Ops instances, your scenario is a pretty straightforward one, however I caution you on the planned means of migrated your legacy 5.x instance to v6.x over the WAN. Doing a data migration over a WAN can become problematic, and network latency can give you some headache in the form of failed data migration and PROLONGED migration durations. You can always give it a shot and see, but be aware of that as a risk that you may end of mitigating to get the data in to v6. For slow WANs, some folks even migrate+re-IP v5 to be at the same data center as the new v6 instance to complete the migration (this results in some downtime in the data for the migration, but in the grand scheme of things it's not the end of the world).

On the objects/metric count of the sizing of the new instance, the reason we look at both object # and metric # is because metrics generate load on the system, however objects do as well. Object have a bit of overhead for vR Ops-metric generations, query overhead/etc on the backend. My suggestion would be to stick with that sizing guidance, and if you indeed have 55k objects you WOULD need a minimum of 6 Large vR Ops nodes.

If you find this or any other answer useful please mark the answer as correct or helpful.
Reply
0 Kudos