Question about vCO architecture

[knight77]

Hello, I am working on rolling out our vCO environment.  We currently have about 18 VC’s mainly in the US and Europe, but are expanding elsewhere in the world really soon.  I know the vCO is bundled yet disabled with each VC and that a single standalone vCO installation can handle more than 10 VC’s. That said what is the commonly accepted architecture for large scale/global vCO installations? If I have a vCO installation in LA, and I deploy a VM to New York, can I select  a template in NYC to deploy or does a template from LA get pushed to NYC?

[cdecanini_]

vCO should be sitting "near" to vCenter, meaning low latency / high bandwith.

To make these vCO talk together you can use the multi node plug-in so a master vCO can start workflows on any of the geos.

To have HA you can cluster each of the vCO server.

[knight77]

Hello, thanks for the reply.  I had a look at this https://www.vmware.com/support/orchestrator/doc/multi-node-plugin-551-release-notes.html When using multi node should the master vCO server be standalone vapp with SQL, IE does need to have more resources than each of the slaves?

[cdecanini_]

The master may run more workflows than each of the slaves because it may run a proxy workflow for each workflow running on the slaves. I have not yet looked into recent releases of multinode to check if the proxy workflow was using any kind of "wait for event" that would eliminate this overhead but this is fine for running up to 300 concurent workflows.

In general it is recommended to install vCO as stand alone (meaning not the one in vCenter) since it is easier to operate (updates / monitoring / backup).

[knight77]

Thank you for the info

[mcfadyenj]

also you should be aware that introduction of multi-node plugin can cause duplicate workflow names to arise.

If you are programmatically accessing the workflow list that can cause issues with duplicate filenames. I found that out first hand after adding the multi-node plugin.

[knight77]

mcfadyenj, can you elaborate on your architecture and how you were using the plug-in?  Avoiding an issue would be handy for a change!