Pipeline won't refresh docker image to latest

Question

I recently updated a docker image used by an existing pipeline. But the pipeline continues to use a older (cached?) image.

Old image sha: af561898ba
New image sha: 4d1f87d6c1

I've tried deleting the cache (not positive I did this correctly, we don't have any caching defined) and manually pulling the latest image in the pipeline yaml:

- docker pull company/project:latest

Which succeeds:
Digest: sha256:4d1f87d6c1f8cec1e816558d176c2202feb3cb3c7a2f79321462e8b1b7f8275eStatus: Downloaded newer image for $DOCKERHUB_USERNAME/project:latest
docker.io/$DOCKERHUB_USERNAME/project:latest But the later steps of the yaml still continue to use the older image:

From the 'Build Setup':
Images used:
build : docker.io/$DOCKERHUB_USERNAME/project@sha256:af561898bae54681f84551a1d4d4bc5aeeaa528e1d4097221dd671d553af6e5d


The new image is available and 100% working, I was able to manually pull it and test it locally outside of the pipe.  I also deleted all previous images on docker, so this must be a cached version via bitbucket some how.

I think I need help a) clearing the docker image cache on bitbucket, or b) forcing pipe to use a specific docker image.  I've already tried:

image:
name: company/project:latest
username: $DOCKERHUB_USERNAME
password: $DOCKERHUB_PASSWORD
email: $DOCKER_IMAGE_REGISTRY_EMAIL

for b) but it has no affect.

Stephen Moon · Answer

Appreciate the information, but unfortunately I wasn't able to get this to work either.

Theodora Boudale · Answer

Hi Stephen,

I would suggest creating a ticket with the support team about this issue. If we have an open support ticket, the engineer working on it will be able to access additional logs and investigate.

You can create a ticket via https://support.atlassian.com/contact/#/, in "What can we help you with?" select "Technical issues and bugs" and then Bitbucket Cloud as product. Please make sure to provide the build URL in the ticket description, so we can check the build logs and additional logs.

Kind regards,
Theodora

Aron Gombas _Midori_ · Answer

I  remember we ran into this a few years ago. Our solution is that we update the Docker image only from the Pipeline with a custom build, which somehow flushes the previously cached Docker image.

It roughly looks like this (with parts removed):

pipelines:
  default:
    ... removed
  tags:
    'docker-wdio-*':
      - step:
          image: atlassian/default-image:3
          name: Update Pipeline Docker image
          caches:
            - docker
          services:
            - docker
          script:
            - cd test/synthetic
            - docker build -t midoriglobal/carchc-wdio:latest .
            - docker login --username $DOCKER_USERNAME --password $DOCKER_PASSWORD
            - docker push midoriglobal/carchc-wdio:latest
  custom:
    wdio-tests:
      - step:
          image: midoriglobal/carchc-wdio:latest
          name: Browser tests
          caches:
            - node
          script:
             .... removed

As you see, you have to run the build "docker-wdio-*" to update the image, which will be used in all future "wdio-tests" builds.

I hope this helps.