Procedure for activating clustering on Atlassian Helm Charts for Confluence

[robert.haskins]

I have had issues converting a test Confluence installation (with no content) from a single node to clustered instance. Is there any documentation that goes over the correct procedure? I have looked through the docs (https://atlassian.github.io/data-center-helm-charts/) and I don't see much if anything that outlines the full procedure for this.

Thanks!

[Jason Allen]

Hello @robert.haskins

We had a similar problem when attempting to scale an existing Atlassian Data Center Confluence deployment from a single pod to multiple pods.  If nothing has changed in the past year, it can't be done unless you answered "Clustered" in lieu of "Single node" in the "Clustering" dialogue during initial setup.  If you answered "Clustered", you can then scale the deployment using 'helm upgrade...' after modifying file "values.yaml" to increase "replicaCount".  If you answered "Single node" you may have to reset and re-deploy the application and all resources particularly the database, which retains a record that the initial setup has been completed.

[robert.haskins]

Hi @Jason Allen 

Thanks for your note! We would have had a very similar path to you (would have gone live with single pod) except that I managed to get clustering working correctly thanks to help from Atlassian support. The only way I could get the clustering process to work was the following process:

1. Initial install with values.yaml/things set as follows:
   1. replicaCount: 1
   2. clustering.enabled: true
   3. ensure license is in kubernetes secret
2. Do install; should skip over license entry and go right to "input content" screen (and enter admin user screen)
   1. Sometimes it prompts for license input at this step
   2. Doesn't seem to cause an issue down the line if it does require the license to be entered
3. Uninstall Confluence
4. Make following setting changes to values.yaml:
   1. replicaCount: 2
   2. "kubectl delete" pv and pvc for original EBS volume
5. Re-deploy Confluence
   1. Comes up to "normal" login screen

A few issues with the docs and dialog screens on this process IMHO:

1. GUI dialog says that you can "upgrade to clustering later" (or something similar) but I have not found a way to do this successfully like you mentioned
2. Unclear what the statement "clear local-home pvc" means (see https://github.com/atlassian/data-center-helm-charts/issues/555); in reality it means "kubectl delete pvc <pvc>" (in addition to doing the same for the pv BTW) on the EBS volume
3. In a single pod setup, the attachments are placed on the (what I understood to be) "disposable" EBS volume; we *almost* went live with this setup. It is unclear from the documentation that in a single node setup, the attachments aren't on the shared-home volume like they are for a clustered installation

[Jason Allen]

Hello @robert.haskins 

I don't know whether or not this is going to show up threaded or not, but it's a reply to your post "below".

Thanks for the note.  I'll give that a try the next time we're working with Atlassian Data Center Confluence.  Maybe something changed in the past year, or maybe deleting the PV and PVC is the "secret sauce" we were missing.