Bitbucket Pipeline steps limit

[jcorzani]

Hi,

I did notice that Bitbucket Pipeline has a 10 steps limit.

It's very easy to reach that limit considering more than one environment with few parallel steps. How do I describe a more complex pipeline that has more than 10 steps and doing it without grouping commands under the script section of a single step?

For example if I have 3 environments to deploy to, each of them needs a manual step just to block the parallel ones so at least 3 steps are wasted:

- step:
    name: Test Environment Deploy
    trigger: manual
    deployment: test
    script:
      - echo "This is a dummy step, just to block the parallel ones"
- parallel:
    - step:
        name: Parallel 1
        script:
          - something
    - step:
        name: Parallel 2
        script:
          - something else

To reach the limit I simply need to have 3 environment like the previous one.

[Philip Hodder]

Hi Jacopo,

Currently Pipelines has a 10 step limit. However, we have an open feature request here to bump the limit, which you should follow: https://bitbucket.org/site/master/issues/16558/increase-multiple-steps-limit

The only feature that I think could currently help, depending on your workflow, is the after-scripts configuration. This will allow you to set up script commands that execute one a given step completes, useful for various teardown operations you may want to do regardless of if a step fails or succeeds.

Out of curiosity, would you be willing to share what you deployment workflow looks like and why you need the manual step blocking parallel steps for a deployment?

Thanks,

Phil

[jcorzani]

Hi Phil,

thanks for your answer. At that time I fixed the problem by creating multiple "custom" entries on the pipeline yml. Instead of pushing the button deploy and having everything in one single screen, I need to go to a specific commit and trigger deploy_to_XYZ. I don't have to do the entire cycle very frequently so that was a good trade-off. This is the solution I've used:

image: openjdk:8
pipelines:
  branches:
    master:
      # deploy-to-qa and deploy-to-live were supposed to be here separated by a manual step
      - step:
          name: Build and Local Test
          script:
            - build and test
  custom:
    deploy-to-qa:
      - step:
          name: QA
          script:
            - pre...
      - parallel:
        - step:
            name: Deploy 1
            script:
              - deploy 1 to QA
        - step:
            name: Deploy 2
            script:
              - deploy 2 to QA
        - step:
            name: Deploy 3 to QA
            script:
              - deploy 3 to QA
      - step:
        name: Final check
        deployment: test
        script:
            - echo "QA Deployed :)"  # This is a one of the dummy steps to set up the property deployment 
    deploy-to-live:
      - step:
          name: LIVE
          caches:
            - gradle
          script:
            - pre...
      - parallel:
        - step:
            name: Deploy 1
            caches:
              - gradle
            script:
              - deploy 1 to LIVE
        - step:
            name: Deploy 2
            caches:
              - gradle
            script:
              - deploy 2 to LIVE
        - step:
            name: Deploy 3 to LIVE
            caches:
              - gradle
            script:
              - deploy 3 to LIVE      
      - step:
          name: Final check
          deployment: production
          script:
            - echo "LIVE Deployed :)" # This is a one of the dummy steps to set up the property deployment 

My problem was making all those steps (I have 3 envs, here are just 2) a single pipeline with a manual step between those instead of triggering deploy_to_X separately. Another problem I had was wasting a step to set up the deployment type "test, production..." because parallel doesn't let you specify it so a dummy step "echo ..." was created.