after_script commands are not run after a multi-stage Dockerfile build with kaniko

Summary

The after_script commands are not run after a build (using kaniko) of a multi-stage Docker image.

It works as expected if this is a single stage Docker image.

Steps to reproduce

The bug can be reproduced with the following simple .gitlab-ci.yml file, the build job will work as expected but not the build_multi_stage one.

.gitlab-ci.yml

stages:
  - build

.build:
  stage: build
  image:
    name: gcr.io/kaniko-project/executor:v1.9.1-debug
    entrypoint: ['']
  script:
    - /kaniko/executor
      --context $CONTEXT
      --dockerfile ${DOCKERFILE_PATH}
      --no-push
  after_script:
    - echo "applesauce"
  variables:
    CONTEXT: "."

build:
  extends: .build
  variables:
    DOCKERFILE_PATH: Dockerfile

build_multi_stage:
  extends: .build
  variables:
    DOCKERFILE_PATH: Dockerfile_multistage

The file Dockerfile:

FROM alpine:3.16

RUN apk upgrade --no-cache

The file Dockerfile_multistage (yes, it's a silly one):

FROM alpine:3.16 AS builder

FROM alpine:3.16

RUN apk upgrade --no-cache

I tried with:

• different version of kaniko (1.9.1, 1.9.0, 1.8.0, 1.7.0)
• different Dockerfiles (as soon as there are multiples stages, it fails to run the after_script)

Actual behavior

The after_script is skipped / not run after my kaniko build of a multi-stage image.

Expected behavior

The after_script should run after my kaniko build of a multi-stage image.

Relevant logs and/or screenshots

• Logs of the job build

• Logs of the job build_multi_stage, where the after_script is not run

Environment description

config.toml contents

listen_address = "0.0.0.0:9252"
[session_server]
  session_timeout = 1800
[[runners]]
  request_concurrency = 20
  url = "<REDACTED>"
  executor = "kubernetes"
  output_limit=10000
  environment = ["K8S_AUTH_KUBECONFIG=~/.kube/config", "FF_GITLAB_REGISTRY_HELPER_IMAGE=1"]
  [runners.custom_build_dir]
  [runners.cache]
    Type = "azure"
    [runners.cache.azure]
      StorageDomain = "blob.core.windows.net"
      AccountName = "<REDACTED>"
      AccountKey = "<REDACTED>"
      ContainerName = "gitlab-cache"
  [runners.kubernetes]
    namespace = "<REDACTED>"
    namespace_overwrite_allowed = ""
    image_pull_secrets = ["<REDACTED>"]
    cpu_request = "200m"
    cpu_request_overwrite_max_allowed = "2"
    memory_request = "200Mi"
    memory_request_overwrite_max_allowed = "4Gi"
    privileged = false
    poll_interval = 5
    poll_timeout = 360
    host = ""
    bearer_token_overwrite_allowed = false
    image = ""
    pull_policy = "if-not-present"
    service_account = ""
    service_account_overwrite_allowed = ""
    pod_annotations_overwrite_allowed = ""
    cpu_limit = "2"
    cpu_limit_overwrite_max_allowed = "4"
    memory_limit = "3Gi"
    memory_limit_overwrite_max_allowed = "8Gi"

Used GitLab Runner version

GitLab Runner version 15.9.1 with k8s executor (GitLab version 15.9.2)

Running with gitlab-runner 15.9.1 (d540b510)
...
Using Kubernetes executor with image gcr.io/kaniko-project/executor:v1.9.1-debug ...

Possible fixes

No fix in sight!