after_script or equivalent in .gitlab-ci.yml

With reference to the docs

Would be useful to have a step that runs after every other step, regardless of whether the build succeeds or not (essentially a tidy up/tear down step)

My specific use case is wanting to remove a directory from a cache after a build on a feature branch (which is essentially caused by this npm issue)

Thoughts?

Do you think that it should be in the same scope as before_script?

Not sure what definition of scope you're using here...?

@myitcv:

before_script:
  - script_run_before_tests

after_script:
  - script_run_after_running_jobs_script

job:
  script: echo Hello World

Added improvement label

Milestone changed to v0.6.0

@ayufan - yes, exactly what I was thinking.

The thing to note being that the after_script would always be run, even, for example, if the before_script fails

Any update on this feature? It sounds very useful in general. I specifically want to cat test.log | grep "failure", i.e. to see what's caused CTest to fail.

@dzhoshkun Not yet. It will most likely happen in GitLab 8.2.

Milestone changed to v0.7.0

@ayufan I know you're busy, just wanted to know, what is the status of this ?

@g13013 We will think about the plan for this feature in the upcoming planning for 8.4 release. Will let you know then.

Milestone changed to v1.1

Milestone changed to v1.2

We will most likely introduce: finally_script. It's more definitive then after_script. Everything in finally_script will be executed also when build is failed or canceled. This will require significant changes to Runner to make it possible, but this is something needed to do proper cleanup of resources.

Added feature proposal label

@ayufan - finally_script is a much better name. Thanks for the update

Fantastic!

@ayufan I like the idea, but I don't like the name. What about calling it post_script? Similar to to the post_receive hook in git https://git-scm.com/book/en/v2/Customizing-Git-Git-Hooks And it is kinda funny since post script is the P.S. in a letter and a computer language for creating vector graphics.

Will this be available for both individual task and for all tasks?

I really want to use this to clean up signing keys, but I dont install them for all tasks, ie test task doesnt need keys

Related: https://gitlab.com/gitlab-org/gitlab-ce/issues/3805

Does finally_script is planned to be used on each job ?

My use case: after each job script (success or fail) I need to perform a task inside the job then process the next job. As I use CI variables, I really need an isolation for each job for finally_script. Am I clear ? :)

I'd be tempted to keep the name as after_script simply so its the same as Travis uses, making it easier to migrate (or use both as i am sure many people do)

MR https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/3771 for GitLab

@ayufan thanks

Ok. This will be named after_script, it makes sense, because we have before_script

It will also be possible to define these two on job level: https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/3772

Implemented in https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/3771, I think we can close this issue. Feel free to reopen if this does not solve problems listed in this issue.

Status changed to closed

@grzesiek @ayufan does this also run on timeout of a build? A big potential customer is asking.

@JobV Yes. This will also run on timeout. The flow is like follow:

1. clone
2. cache restore / artifacts restore
3. before_script + script
4. after_script
5. cache save / artifacts save

If we timeout or cancel on before_script + script we will also run after_script.

Can after_script timeout?

Thanks @ayufan

@mcfedr Yes, the after_script will have timeout. Not sure where will it be configured, but most likely we will limit it to some arbitrary value (most likely 5 or 10 minutes) that will later be configurable.

mentioned in merge request gitlab-com/www-gitlab-com!2018 (merged)

Is there any way to know the status of the build in after_script? Maybe some environmental variable indicating the status, but I couldn't find anything yet?

@rogosz.marek I don't believe there is, but that's a good idea. I think there might even be an issue for it already. If not, can you create a new feature request with your use-case?

@rogosz.marek Currently not. Currently we have everything to make this possible! Please create a separate issue.

Was there an issue created ?

I would like to see the addition of a property for after_script, the same used for stages

after_script:
     when: on_failure
     script:
           - execute clean up 

our use case is

stages:
    - build_cloud_cluster
    - test_cluster
    - cleanup_cluster

build_cloud_cluster:
    stage: build_cloud_cluster
    script: 
         - make cluster
    after_script:
        when: on_failure
        script:
              - save log as artifact
              - make clean

...
...
       

Right now we can use a separate Stage to handle failure in the previous stage, but it becomes messy in the pipeline view when you end up with a clean up stage for each stages.

https://gitlab.com/gitlab-org/gitlab-ci-multi-runner/issues/1538

Anyone know if after_script is supposed to run even if you cancel a job? It states above that it should but with testing seems like it's not working.

https://gitlab.com/gitlab-org/gitlab-ce/issues/20727 found it - not supported (or at least it's unclear for multiple people)

mentioned in issue #2716