K8SSAND-59 ⁃ How to get multi DC environment created?

[navjyotnishant]

Type of question

How to perform a particular operation

What did you do?
Tried to spin up K8ssandra 2 DC with 3 node in each.

cassandra:
  version: "4.0.0"
  cassandraLibDirVolume:
    storageClass: local-path
    size: 5Gi
  allowMultipleNodesPerWorker: true
  heap:
   size: 1G
   newGenSize: 1G
  resources:
    requests:
      cpu: 1000m
      memory: 2Gi
    limits:
      cpu: 1000m
      memory: 2Gi
  datacenters:
  - name: k8s-dc1
    size: 3
    racks:
    - name: default
  - name: k8s-dc2
    size: 3
    racks:
    - name: default
kube-prometheus-stack:
  grafana:
  adminUser: admin
  adminPassword: admin
stargate:
  enabled: true
  replicas: 1
  heapMB: 256
  cpuReqMillicores: 200
  cpuLimMillicores: 1000

helm install -f k8ssandra.yaml k8ssandra k8ssandra/k8ssandra

Above command only create k8s-dc1 DC and corresponding pods for 3 nodes.

Did you expect to see some different?
I would expect 2 DC to be created with 3 node in each DC.

Environment
DEV

• Helm charts version
  k8ssandra-1.0.0

┆Issue is synchronized with this Jira Task by Unito
┆friendlyId: K8SSAND-59
┆priority: Medium

[jsanda]

Hi @navjyotnishant

k8ssandra does not yet support multi-DC clusters. See #524.

Note that cass-operator does however already have support for multi-DC. It just is not yet exposed in k8ssandra.

[navjyotnishant]

@jsanda

Thanks for revert. Do we know how soon we can get this feature included? Cassandra with single DC is not a production choice for most of the people. What we can do to priorities this?

[jsanda]

Great questions 😄

Do we know how soon we can get this feature included?

My personal preference would be to address this in 2.0. The short version is, I don't want to implement multi-DC support in the Helm charts. See #485 for some background.

What we can do to priorities this?

Comment on and up vote #524. More generally speaking the more involved you can be with the project the better!

In terms of multi-DC support, I think of two high level scenarios - 1) DCs running in the same k8s cluster and 2) DCs running in different k8s clusters.

Do you have more interest in one of those scenarios?

[jsanda]

@navjyotnishant

Can we go ahead and close this?

[navjyotnishant]

@jsanda I am ok if you would like to track this via #524.

To answer your question I am ok with either of the options but preferably independent k8s cluster in each DC.

[jsanda]

With #548 you can set up multiple DCs running in different k8s clusters. There are a couple caveats though. First, pods in both clusters need to be reachable by their IP addresses. Secondly, there is no centralized installation, configuration, management. You will need to run helm install in each k8s cluster.