vROps 6.0.x migration project

Essendon

Hi guys, has anyone done a migration or a fresh install of vROps 6.0.x yet?

I have a pretty complex migration at hand with multiple vCenter Server machines spread across continents. Nice project for sure but a huge headache too. I'm looking into the use of Remote Collectors to send data to the Master Node:

1. One 5.8.3 instance monitoring 9 vCenter machines, this machine is massive - 24 vCPUs and a ton of RAM, burning about 10TB of disk. These vCenter machines are spread across datacenters (with high speed links) on various platforms (Vblock, HP blades etc)

2. One 5.8.3 instance monitoring 5 vCenter machines in another continent. This one's smaller at 16 vCPUs and 48GB RAM chomping about 6TB or something.

Has anyone done a migration of this scale. I did look at the 'data nodes' but these are meant either for HA purposes or for load distribution usually within a datacenter (high speed connection to the master node is a mandatory requirement). Looks like Remote Collectors are the only option, but these things dont chew through any data locally and just send the data across as is. Generally used for navigating across firewalls.

The goal was to have a single point of monitoring but it looks like there'll have to be 2 points of management, one in Australia and the other in the second continent (just like there is now).

Have ideas?

Deathmage

I have vROps 6 at home. Been beta testing 6 for a while now. But it's not a deployment that is even close to what you're doing. I just did it to mess around with it.

My only concern is can the one site of monitoring compensate for latency of the other locations? (i presume your going for a hub-n-spoke topology) ....going to presume a datacenter in Melbourne is going to have latency with a datacenter in NYC just because of the distance.

Essendon

Good point about the latency between sites. For data nodes - there needs to be a high speed connection between these nodes and data nodes, essentially these machines will need to be in the same datacenter. For the remote collector nodes, they have smarts built into them, they buffer data in memory till there's memory pressure on the VM at which time they'll send data across to the master node. It's estimated for about 1 million metrics, an RC node will send about 300MB/hour to the master node.

For everyone's information, vROps 6.0.1 is out, one stand-out feature for me is how it handles extraneous metrics. Extraneous metrics are set to disabled out of the box now, so the installation wont burn as much disk as it did with vROps 6.0.0. Did you know they put in this feature because of a bug request I made? When I installed 6.0.0, it went through about 500GB inside 3 days. VMware Support asked me to file a bug request and thankfully it's been fixed.

So if you are thinking of migrating to 6.0, install 6.0.0 then immediately go to 6.0.1 to avoid those disk usage issues.

dave330i

My boss keeps asking me to learn vROps, but there's a limit to how much stuff I can learn.

GL on your design.

Essendon

Learn it Dave, you wont take long to pick up and it's not the hardest thing to design either.

dave330i

Essendon wrote: »

Learn it Dave, you wont take long to pick up and it's not the hardest thing to design either.

If I learn vROps more than the basic install, I'd be traveling non-stop with BCDR, NSX & vROps engagements and my wife would kill me.

JBrown

dave330i wrote: »

If I learn vROps more than the basic install, I'd be traveling non-stop with BCDR, NSX & vROps engagements and my wife would kill me.

Take your wife with you, if engagements are 3 days or more. Hotel/ ticket for you covered already anyways, just get her a ticket. She gets to see now countries/ and travel around, and gets to spend time wth you.

Lexluethar

We migrated but not that scale. We have around 500 vms and about three dozen hosts. We used vCOPS (still use) 5.7.2 in our environment and moving towards vROPS 6.0. The roadmap that was laid out for us by VMware was to standup a vROPS instance, so deploy the ova, configure accordingly with the AD accounts, what level of monitoring, ect then allow that collector to continue collecting for your retention period in your vCOPS environment. For us 30 days - now after 30 days we can decom the vCOPS instance and move over to vROPS.

Something to keep in mind to is some of the new features that are coming down the pip with vSphere 6.0. I realize no large company will move quickly to it, but some of the enhanced features like being able to have across data center vCenter's (IE higher latency) may allow you to consolidate some of your vCenter instances.

Essendon

@Lexluether - thanks for the input, yeah that's how I've migrated our lab instance too. Retention period's 6 months for us, hence the heavy disk consumption. The migration of data from the older version didnt go too well when I tried a migration and someone at VMware told me to stand a vROps 6.0 instance in parallel.

dave330i

JBrown wrote: »

Take your wife with you, if engagements are 3 days or more. Hotel/ ticket for you covered already anyways, just get her a ticket. She gets to see now countries/ and travel around, and gets to spend time wth you.

She has her own 6 figure career to deal with.

Lexluethar

I can only imagine migrating the data would be a pain considering how much you have in place. Not sure if resources are a constraint for you but could you just leave the 5x environment up and use it for past performance viewing?