Confluence Data Center deployment on AWS EKS

Gal Bakal

We are trying to deploy Confluence through data center cluster by using helm charts in Kubernetes cluster.

Below are some of the concerns we have regarding the whole setup:

What are the pros and cons of using ebs for local-home and efs for shared-home. We noticed as per Atlassian's recommendation
(https://atlassian.github.io/data-center-helm-charts/userguide/PREREQUISITES/), the document says to use efs for shared home and ebs for local-home, but with this set up it's difficult to scale as replicas are using different EBS volume. We want to understand what kind of data is being stored in Local-Home and can we use EFS for Local-Home ?

We would like to understand the difference between Standalone and Clustered deployment options that Confluence offers. Can we achieve multi-pod feature and high availability within Standalone deployment (considering we have HA set up on our side)?

Sagar Mahajan_d8sqh7

Hi @Gal Bakal Let me explain you what I learn from our data center setup on Openshift platform.

 

We want to understand what kind of data is being stored in Local-Home and can we use EFS for Local-Home?

Local-Home - Confluence local contains the dynamic files like index folder (index file of Confluence), Logs files, database config file, etc.

Shared-Home - Mostly the static files like Attachments, installed-plugins, export, restore, script folders, etc

This is just a high level directories which are part of local-home and shared-home.

I would suggest use the local instance storage like EBS for local-home as the indexes file are stored here and disk I/O should be good so that indexs should be updated from the database whenever we search for any document in Confluence and that details are fetched from the database and indexes will be get updated in local home, also these indexes are replicated to other nodes in cluster so the read/write operations should be fast.

Initially we used the EFS storage by creating new access point for the local-home but after initial performance test we are getting lot of index replication health check failed warnings, then after some analysis we moved to local storage which is EBS in our case as we are using EC2 instance for deploying our applications (master, worker nodes) and solve our issue after testing disk I/O

We would like to understand the difference between Standalone and Clustered deployment options that Confluence offers.

The standalone deployment is same as that of running a single node or which was previously as Server version.

Not require any shared-home to be configured, load balancer etc, just update your Confluence server license with the data-center one and your Confluence instance will be converted to datacenter as for standalone installation only license change is fine, no changes on application architecture etc. updating the license will automatically include all the application features which are available for Datacenter

Clustered deployment is the one where you have at least 2 or more nodes, which require shared-home, load balancer to balance the traffic between nodes, auto-scaling etc.

Can we achieve multi-pod feature and high availability within Standalone deployment

You will not able to achieve the high availability, fail over, Performance at Scale as your application is running on single node and standalone deployment does not have options for multi-pod deployment.

considering we have HA set up on our side

Not fully understand what this mean but if you talking about you setup a DR instance in ant other availability zone or region then also it take some time to update the DNS and other config to redirect the traffic to DR instance, also I think it will not be a cost efficient as you require the application instance and database instance to be running all the time and active:active replication between your production environment to DR environment.

 

I know this is a lot, but let me know in case any questions or concerns.