I have an application in GCP, the packer flow is like this:
GCP load balancer ---- proxy server (nginx in and incus instance) ---- application server (apache+tomcats in incus instance).
It’s a POC of a project and works well.
Now, I want to scale, meaning there will be a bunch of application servers and the proxy server will be the frontend all of them. I want to distribute the load between multiple GCP instances, so an instance would host the proxy Incus instance and multiple (larger) instances would host the application servers.
I hope the plan is clear:)
My question is what is the best or what are the possible Incus deployment strategies? I am looking at Incus clustering in the docs, but unsure if that is possible to deploy in GCP. Is it so?
If possible, I would like to be able to migrate Incus instances between GCP instances.
Clustering works fine on GCP. We actually used to use GCP for clustering benchmarking
The networking is likely where things may get a bit weird on GCP when you have multiple servers.
You certainly set up your own network overlay with OVN and your instances will be able to talk to each other regardless of what server they’re running on.
The issue is going to be with handling ingress into the OVN network. You may need to have one server that’s effectively the only one running the routers and so the only one that handles traffic to your public IPs. Incus lets you do that through cluster roles, but you will effectively find yourself with a single point of failure there.
I don’t know if anyone else has comments on how they’re handling that on GCP.
Thanks. I had the feeling that I needed to go this way:)
Is there a more or less step-by-step guide for it? Or should I just follow the standard clustering documentation?
Frankly, I have never used OVN.
