Docker Entrypoint setting ignored during update

[Khoulaiz]

Summary

When updating an existing docker image with a new version of the image, the ENTRYPOINT settings from the updated image is ignored and the old ENTRYPOINT setting is applied (according to the output of the cf push command).

I changed the ENTRYPOINT in my image from
ENTRYPOINT ["/go/bin/imaginary", "-cors", "-concurrency", "20"]
to
ENTRYPOINT ["/go/bin/imaginary", "-cors", "-concurrency", "20", "-enable-url-source"]

Expected Result

Entrypoint for the docker image in cf should be
ENTRYPOINT ["/go/bin/imaginary", "-cors", "-concurrency", "20", "-enable-url-source"]

Actual Result

The image was updated but the entrypoint remained to
ENTRYPOINT ["/go/bin/imaginary", "-cors", "-concurrency", "20"]

Context

I am using this microservice via the docker image. Updating the image without interruption would be important once we are going in production.

Steps to Reproduce

1. Create a docker image with entrypoint
2. deploy it
3. change the entrypoint by adding another cmd line parameter
4. push it again. Update will be recognized but change will be ignored according to cf push output

[cf-gitbot]

We have created an issue in Pivotal Tracker to manage this:

https://www.pivotaltracker.com/story/show/159065575

The labels on this github issue will be updated when the story is started.

[jvshahid]

Hi @Khoulaiz sorry about the delay in getting back to you. This looks like a CAPI concern since the CC is the component responsible for sending the entrypoint command to Diego in the LRP definition. I am going to ping the Capi engineers to look into this issue, in the meantime I think you can work around this issue by restaging the application.

[jvshahid]

ping @zrob @Gerg

[tcdowney]

Hi @Khoulaiz, I'm a dev on the CAPI (Cloud Controller) team and we have a story in our backlog to look into this:

https://www.pivotaltracker.com/story/show/159065575

[elenasharma]

Hi @Khoulaiz, my pair @goonzoid and I looked into this for you. We were able to reproduce the behavior you are seeing - looking at our logic in CAPI, this is a known limitation for Docker apps. We'd like to better document the functionality so that other users do not run into the same unexpected behavior that you did, and have filed a story to do so here: https://www.pivotaltracker.com/story/show/159417013.

For now, there are a couple of ways you can ensure that updates to your Docker images make their way to your app.

In the short term you can restage your application, as suggested in an above comment.

Long term, we recommend that you tag changes to your Docker image and push with that tag name so that Cloud Foundry registers an update to your Docker image. If you update your Docker image and those changes do not take effect with a cf push, there's a high likelihood that other changes in your environment runtime will cause your changes to go into effect unexpectedly.

We're going to go ahead and close this issue. Let us know if you have any other questions.