Skip to content

Latest commit

 

History

History
60 lines (43 loc) · 3.49 KB

configuration.md

File metadata and controls

60 lines (43 loc) · 3.49 KB

Configuration for Confluence

High Availability

Confluence Clustering has been functional since v8.5.16-uds.0 released October 29th, 2024.

1. Enable Clustering

To enable clustering, set the following values in your UDS bundle.yaml:

packages:
  - name: confluence
    overrides:
      uds-confluence-config:
        uds-confluence-config:
          values:
            - path: clustering.enabled
              value: true
      confluence:
        confluence:
          values:
            # Below are values to enable clustering
            - path: confluence.hazelcastService.enabled
              value: true # Set to true to enable clustering
            - path: confluence.clustering.enabled
              value: true # Set to true to enable clustering
          variables: 
            - name: CONFLUENCE_REPLICAS
              path: confluence.replicaCount
              default: 1  # Do not increase this until step 3 below

Enabling clustering in the uds-confluence-config chart causes additional network policies/rules to be created providing things like cookie-based session affinity through an Istio Destination Rule and permission for the Confluence pods to query the KubeAPI so they can locate and talk to each-other.

You may also need to select a different storage class for the Confluence PVCs to enable multiple read/write access, as every pod will connect to a shared PVC. This settings is at the path: volumes.sharedHome.persistentVolumeClaim.storageClassName.

2. Deploy Confluence

If this is a new Confluence instance, install with just one replica, and complete the setup wizard before moving on to the next step. Enabling clustering and getting more than one replica running before the setup wizard has completed will break the instance.

If this is not a new confluence instance, we still recommend waiting to scale until you've confirmed your pre-existing single confluence instance has stabilized following the upgrade.

3. Scale Confluence, one replica at a time

It is crucial that only one replica is added to confluence at a time, and that those replicas are turned off highest-ordinal-replica first when performing maintenance.

  1. Increment the value of the variable CONFLUENCE_REPLICAS in your UDS configuration file.
  2. Deploy your bundle, (you may use add the -p confluence flag to save time).
  3. Return to step 1 above until you've reached the desired number of replicas.

You can replace steps 1 and 2 with the command kubectl scale -n confluence statefulset confluence --replicas X. Still scale it one replica at a time, wait until that replica is joined into the confluence node cluster (watch the logs and make a good guess), and add another replica. If you take this approach, be sure to update the CONFLUENCE_REPLICAS value in your uds configuration file.

Upgrading Clustered Confluence

  1. Scale Confluence down to 1 replica.
  2. Deploy the bundle update.
  3. Scale Confluence back up as described above.

If you skip step 1, Kubernetes will replace the highest-ordinal pod, then the next lowest, and so on down. This will cause higher ordered pods to be running while lower-ordered pods are dying and updating. This might work, but is not tested, and goes contrary to this author's recollection of the Confluence documentation. Confluence tends to care a lot that nodes turn on in the order 0, 1, 2, 3 and turn off in the opposite order.