Page MenuHomePhabricator

Kubernetes prod and staging share the same cluster key in hiera
Closed, ResolvedPublic

Description

I stumbled upon this fact as I was looking at the etcd dashboard. It filters by label "cluster" which is generated puppet via prometheus::class_config from hiera key cluster: which we set to kubernetes for prod and staging:

hieradata/role/common/etcd/kubernetes.yaml:cluster: kubernetes
hieradata/role/common/etcd/networking.yaml:cluster: kubernetes
hieradata/role/common/etcd/v3/kubernetes.yaml:cluster: kubernetes
hieradata/role/common/etcd/v3/kubernetes/staging.yaml:cluster: kubernetes
hieradata/role/common/kubernetes/master.yaml:cluster: kubernetes
hieradata/role/common/kubernetes/staging/master.yaml:cluster: kubernetes
hieradata/role/common/kubernetes/staging/worker.yaml:cluster: kubernetes
hieradata/role/common/kubernetes/worker.yaml:cluster: kubernetes

We should probably use kubernetes-staging for staging kubernetes and corresponding etcd, but I'm unsure about the consequences a change like this would have.

Event Timeline

JMeybohm triaged this task as Medium priority.

We should probably use kubernetes-staging for staging kubernetes and corresponding etcd, but I'm unsure about the consequences a change like this would have.

+1. It's been bugging me for a while, but it was always a very low priority since I never really had a strong incentive to "fix it".

As far as consequences, I don't think many. We 'll need to create the icinga groups to not break icinga and we 'll lose some history in dashboards in grafana that filter on the cluster. Maybe some budget related calculation spreadsheets as well. But nothing of that sounds painful.

Change 759470 had a related patch set uploaded (by JMeybohm; author: JMeybohm):

[operations/puppet@production] Create a new wikimedia_cluster: kubernetes-staging

https://gerrit.wikimedia.org/r/759470

Change 759470 merged by JMeybohm:

[operations/puppet@production] Create a new wikimedia_cluster: kubernetes-staging

https://gerrit.wikimedia.org/r/759470

Mentioned in SAL (#wikimedia-operations) [2022-02-03T13:27:16Z] <jayme> moved kubernetes staging master,nodes,etcd from wikimedia_cluster "kubernetes" to "kubernetes-staging" - T273866