Skip to main content

Migrating Bitbucket to Kubernetes

— Kyungmin Park

This document explains the process of migrating an existing Bitbucket installation to Kubernetes.

💡
Bitbucket 8.9.26 → 9.4.7

Pre-configured resources

Helm Chart

Version: v3.17.3(GoVersion: go1.23.7)

Add the Helm chart repository

helm repo add atlassian-data-center \
  https://atlassian.github.io/data-center-helm-charts

Check Atlassian product version with chart version

helm search repo atlassian-data-center/bitbucket

Get values.yaml command

helm show values altassian-data-center/bitbucket --version 1.22.0 > values.yaml

Editing values.yaml

Modify according to your configuration.

Create namespace

kubectl create ns atlassian

Deploy helm chart

Bitbucket version: 8.9.26

helm install bitbucket \
 atlassian-data-center/bitbucket \
 -f values.yaml \
 -n atlassian \
 --version 1.22.0

Edit Bitbucket statefulset scale

kubectl scale statefulset --replicas 0 -n atlassian bitbucket
💡
"Why is the StatefulSet's replicas set to 0 immediately after deploying the Bitbucket Helm chart?"
To prepare for migration by creating an empty Persistent Volume (PV), similar to the guides for Jira, Confluence, and Crowd.
The initial setup screen does not appear, so the replicas are immediately set to 0.

Data Migration

Process for importing existing application data and using it in the existing environment

Migration strategy

  1. Bitbucket deploy(replicas=1 → replicas=0) with empty data(application-data, shared-home)
  2. Deploying a BusyBox pod with Bitbucket's local-home and shared-home mounted (I used a Debian image instead of BusyBox, as I’m more familiar with it)
  3. Get the application-data directory from the as-is Bitbucket instance and rsync it into the Bitbucket volume using a BusyBox pod
  4. Get the shared-home directory from the as-is Bitbucket instance and rsync it into the Bitbucket volume using a BusyBox pod
  5. Rescale Bitbucket statefulset(replicas=0 → replicas=1)
  6. Check Bitbucket data
  7. Rescale Bitbucket statefulset(final replicasCount)

Upgrade

Upgrade strategy

  1. replicas 0(shutdown all Bitbucket pod)
  2. Set image.tag
  3. Helm chart upgrade
  4. Bitbucket statefulset scaling

Shutdown all Bitbucket pod

kubectl scale statefulset --replicas 0 -n atlassian bitbucket

For a Rolling upgrade, adjusting the replicas is not necessary.

Set image.tag

If you previously set image.tag but the APP VERSION matches the version you want to upgrade to, proceed with this step.

image:

  # -- The Docker Crowd Docker image to use
  # https://hub.docker.com/r/atlassian/crowd
  #
  repository: atlassian/crowd

  # -- Image pull policy
  #
  pullPolicy: IfNotPresent

  # -- The docker image tag to be used. Defaults to appVersion in Chart.yaml
  #
  tag: ""

values.yaml

Helm chart

Bitbucket version: 8.9.26 → 9.4.7

helm upgrade bitbucket \
 atlassian-data-center/bitbucket \
 --version 2.0.2 \
 -f values.yaml \
 -n atlassian
🗒️
During helm upgrade, pods are created based on the replicaCount in values.yaml, and the first pod performs the upgrade task.
The previous --replicas 0 setting is ignored.

Using Bitbucket SSH with Kubernetes

In my case, I configured Bitbucket SSH access using OPNSense port forwarding combined with MetalLB.

The values.yaml file provides options to enable SSH access to Bitbucket.

  sshService:

    # -- Set to 'true' if an additional SSH Service should be created
    #
    enabled: true

    # -- Port to expose the SSH service on.
    #
    port: 22

    # -- The hostname of the SSH service. If set, it'll be used to configure the SSH base URL for the application.
    #
    host:

    # -- SSH Service type
    #
    type: LoadBalancer

    # -- Use specific loadBalancerIP. Only applies to service type LoadBalancer.
    #
    loadBalancerIP: <The LoadBalancer IP you will assign>
Successfully performed a git clone over SSH.

Reference

Products upgrade - Atlassian DC Helm Charts
logo