About

GitHub action to implement mutual exclusion within your workflow. It prevents parallel execution of "protected" steps in your workflow by using a kubernetes secret as a distributed lock.

Example

In the following example we protect the "Deploy" step from concurrent execution by other runs of the workflow.

on:
  push:
    branches:
      - master
name: Deployer
jobs:
  deploy-plz:
    runs-on: ubuntu-18.04
    steps:
    - name: Checkout
      uses: actions/checkout@v2
    - name: Lock workflow
      uses: directangular/k8s-lock-action@v1
      with:
        kube_config_data: ${{ secrets.KUBE_CONFIG_DATA }}
        lock_name: "my-deploy-lock"
    - name: Deploy
      uses: ./actions/deployer/

Note that you'll need to make any necessary workspace preparations for proper operation of your kubeconfig. For example, with EKS you would need to configure your AWS credentials prior to k8s-lock-action since your kubeconfig likely uses the aws eks command to authenticate. The image this action runs with includes the awscli package, which will pick up authentication from aws-actions/configure-aws-credentials or similar. Here's the above example with AWS credential configuration added:

    - name: Checkout
      uses: actions/checkout@v2
    - name: Configure AWS Credentials
      uses: aws-actions/configure-aws-credentials@v1
      with:
        aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
        aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
        aws-region: us-west-2
    - name: Lock workflow
      uses: directangular/k8s-lock-action@v1
      with:
        kube_config_data: ${{ secrets.KUBE_CONFIG_DATA }}
        lock_name: "my-deploy-lock"
    - name: Deploy
      uses: ./actions/deployer/

Inputs

kube_config_data

Required. Use base64 < ~/.kube/config (or similar) to generate.

lock_name

Required. Must be unique across the repository where this action is used (otherwise unrelated actions might block each other).

kube_context

Optional. The context from the provided kubeconfig that should be used. This string is simply passed to kubectl use-context.

max_attempts

Optional. How many times we try to grab the lock before bailing. Defaults to 10000.

How it works

This action abuses kubernetes secrets by using them as a distributed lock. The value of the secret is the build number that is currently allowed to execute. During the post action cleanup phase the value of the secret is incremented, allowing the next build to execute.

Limitations

Stuck locks

If the secret gets messed up somehow the task will block until it hits max_attempts, and then will fail. You can unblock it by overwriting the secret manually:

kubectl create secret generic <secret_name> --from-literal=next_build=<next_build> --dry-run -o yaml | kubectl replace -f -

(The exact command is also displayed at the beginning of the lock process for easy reference.)

Supported k8s environments

Currently only tested on EKS. Test results and patches for other environments are welcome.

Re-runs

GitHub does not increment the GITHUB_RUN_NUMBER when you re-run a workflow. Therefore, re-runs will never successfully grab the lock since the lock value increases monotonically with each build.

The bottom line is that you should avoid re-running your builds. If you really want to re-run a build you'll just need to intervene as per the "Stuck locks" section above.