chore(kuma-cp) deploy many instances using HELM #1208
Conversation
Signed-off-by: Jakub Dyszkiewicz <[email protected]>
There was a problem hiding this comment.
Looks like a really nice change to me. We may want to investigate some default HPA policies if this is a common use case, I think @tharun208 had some ideas in the initial Helm chart implementation.
Edit: just saw your issue #1207, looks like we're on the same page. Happy to take a stab at that.
Assigned it to you for visibility. |
| @@ -16,6 +16,8 @@ controlPlane: | |||
|
|
|||
| kdsGlobalAddress: "" | |||
|
|
|||
| replicas: 1 | |||
There was a problem hiding this comment.
I know why it is one by default, but in reality - do we advise people to run a single copy of the CP?
There was a problem hiding this comment.
On production? No, you should run at least 2 for HA (and probably with a PodAffinity settings that two of them are not spawned on the same Node). Ideally, this will be controlled by Horizontal Pod Autoscaler, hence the issue I created.
So what's your suggestion for the default value?
There was a problem hiding this comment.
OK lets go for the HPA apraoch and leave this one as 1.
Signed-off-by: Jakub Dyszkiewicz <[email protected]>
Signed-off-by: Jakub Dyszkiewicz <[email protected]>
Signed-off-by: Jakub Dyszkiewicz <[email protected]> (cherry picked from commit fd0bf9f) # Conflicts: # app/kumactl/pkg/install/k8s/control-plane/helmtemplates_vfsdata.go
Signed-off-by: Jakub Dyszkiewicz <[email protected]>
Signed-off-by: Jakub Dyszkiewicz <[email protected]>
Summary
Add capability to deploy many instances of Kuma by parametrizing HELM. Kuma is meant to be scaled horizontally. I also parametrized Ingress deployment.
Full changelog
RestartKumafor E2E since it's not used anywhereDocumentation