[BUG] compose up --wait exits 1 on init containers successfully completing

[nmguse-bighealth]

Description

When using docker compose up --wait, where I have an init container that populates some data and then exits (0), docker compose will exit 1.

If any other container depends on the init container finishing, it will exit properly, as described/exampled in this PR: https://github.com/docker/compose/pull/9572. In my situation however, there is no container that runs a service that depends on the init container finishing.

Steps To Reproduce

docker-compose.yml example:

version: '3'

services:
  postgres:
    image: postgres:14-alpine
    environment:
      POSTGRES_DB: postgres
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: postgres
    healthcheck:
      test: ['CMD', 'pg_isready']

  postgres_setup:
    image: alpine
    depends_on:
      postgres:
        condition: service_healthy
    restart: "no"
    command: pwd

Run docker compose/print exit code:

$ docker compose up --wait
[+] Running 2/3
 ✔ Network test_default             Created                                                                                                                                                    0.1s
 ✔ Container test-postgres-1        Healthy                                                                                                                                                   31.9s
 ⠿ Container test-postgres_setup-1  Waiting                                                                                                                                                   31.9s
container test-postgres_setup-1 exited (0)

$ echo $?
1

Compose Version

Docker Compose version v2.17.3

Docker Environment

Client:
 Context:    default
 Debug Mode: false
 Plugins:
  buildx: Docker Buildx (Docker Inc.)
    Version:  v0.10.4
    Path:     /Users/nathan/.docker/cli-plugins/docker-buildx
  compose: Docker Compose (Docker Inc.)
    Version:  v2.17.3
    Path:     /Users/nathan/.docker/cli-plugins/docker-compose
  dev: Docker Dev Environments (Docker Inc.)
    Version:  v0.1.0
    Path:     /Users/nathan/.docker/cli-plugins/docker-dev
  extension: Manages Docker extensions (Docker Inc.)
    Version:  v0.2.19
    Path:     /Users/nathan/.docker/cli-plugins/docker-extension
  init: Creates Docker-related starter files for your project (Docker Inc.)
    Version:  v0.1.0-beta.4
    Path:     /Users/nathan/.docker/cli-plugins/docker-init
  sbom: View the packaged-based Software Bill Of Materials (SBOM) for an image (Anchore Inc.)
    Version:  0.6.0
    Path:     /Users/nathan/.docker/cli-plugins/docker-sbom
  scan: Docker Scan (Docker Inc.)
    Version:  v0.26.0
    Path:     /Users/nathan/.docker/cli-plugins/docker-scan
  scout: Command line tool for Docker Scout (Docker Inc.)
    Version:  v0.10.0
    Path:     /Users/nathan/.docker/cli-plugins/docker-scout

Server:
 Containers: 7
  Running: 4
  Paused: 0
  Stopped: 3
 Images: 69
 Server Version: 23.0.5
 Storage Driver: overlay2
  Backing Filesystem: extfs
  Supports d_type: true
  Using metacopy: false
  Native Overlay Diff: true
  userxattr: false
 Logging Driver: json-file
 Cgroup Driver: cgroupfs
 Cgroup Version: 2
 Plugins:
  Volume: local
  Network: bridge host ipvlan macvlan null overlay
  Log: awslogs fluentd gcplogs gelf journald json-file local logentries splunk syslog
 Swarm: inactive
 Runtimes: io.containerd.runc.v2 runc
 Default Runtime: runc
 Init Binary: docker-init
 containerd version: 2806fc1057397dbaeefbea0e4e17bddfbd388f38
 runc version: v1.1.5-0-gf19387a
 init version: de40ad0
 Security Options:
  seccomp
   Profile: builtin
  cgroupns
 Kernel Version: 5.15.49-linuxkit
 Operating System: Docker Desktop
 OSType: linux
 Architecture: x86_64
 CPUs: 4
 Total Memory: 5.804GiB
 Name: docker-desktop
 ID: dc980d43-b287-4b8d-90b1-992be4c7b457
 Docker Root Dir: /var/lib/docker
 Debug Mode: false
 HTTP Proxy: http.docker.internal:3128
 HTTPS Proxy: http.docker.internal:3128
 No Proxy: hubproxy.docker.internal
 Registry: https://index.docker.io/v1/
 Experimental: false
 Insecure Registries:
  hubproxy.docker.internal:5555
  127.0.0.0/8
 Live Restore Enabled: false

Anything else?

No response

[ndeloof]

There's unfortunately no way to declare your setup service isn't actually a service, and termination is expected.

[mvejmelk]

Hit this as well - does anyone have some workaround, please?

[olek-p]

Hit this as well - does anyone have some workaround, please?

I'm using this:

docker-compose -f docker-compose.yml up --wait || true

[mvejmelk]

docker-compose -f docker-compose.yml up --wait || true

We use the docker compose as part of test setup - the or-true approach will effectively void ANY kind of error coming from the stack (not only the "harmless one").

[olek-p]

docker-compose -f docker-compose.yml up --wait || true

We use the docker compose as part of test setup - the or-true approach will effectively void ANY kind of error coming from the stack (not only the "harmless one").

Yes, it will. It will also make your test setup complete without errors (until this issue is fixed). The choice is yours :)

[douardda]

There's unfortunately no way to declare your setup service isn't actually a service, and termination is expected.

what about just using the return code of the "pseudo service" process? if it's 0, then do not consider it as an error by the compose up --wait command.

[luciangabor]

While this is not to argue that --wait couldn't/shouldn't take into account the exit code of services that exit (with the "complications" given by a restart policy taken into account), and with the observation that it's kinda' like that already:

$ docker compose -f - up --wait <<'EOF'
services:
  sleepy:
    image: alpine
    command: sleep 5
EOF
[+] Running 1/1
 ✔ Container waitish-sleepy-1  Healthy

$ echo $?
0

... there is also this page of documentation: https://docs.docker.com/compose/profiles/#auto-starting-profiles-and-dependency-resolution

Well, noting that ... there is an unfortunate limitation - run can only run one service - and ... extrapolating a bit, here's an example of what works now:

# compose.yaml

services:
  postgres:
    image: postgres
    environment:
     POSTGRES_PASSWORD: password
    healthcheck:
     test: pg_isready
     start_interval: 1s
     start_period: 30s

  psql:
    profiles:
      - .init
    image: postgres
    depends_on:
      postgres:
        condition: service_healthy
    environment:
     PGUSER: postgres
     PGDATABASE: postgres
     PGPASSWORD: password
    entrypoint: psql -h postgres
    command: -c \\conninfo

  init:
    profiles:
      - .init
    build:
      dockerfile_inline: FROM scratch
    init: true
    entrypoint: /sbin/docker-init --version
    depends_on:
      psql:
        condition: service_completed_successfully

$ docker compose run init
[+] Creating 3/3
 ✔ Network init_default       Created                                                                                                                                                     0.1s 
 ✔ Container init-postgres-1  Created                                                                                                                                                     0.1s 
 ✔ Container init-psql-1      Created                                                                                                                                                     0.1s 
[+] Running 2/2
 ✔ Container init-postgres-1  Healthy                                                                                                                                                     5.8s 
 ✔ Container init-psql-1      Started                                                                                                                                                     0.3s 
tini version 0.19.0 - git.de40ad0

$ echo $?
0

# postgres is started because of the init -> psql -> postgres depends-ons
# and init could depend on a number of other services/"tasks"/one-offs
# the alternative would be to run each needed "task"/one-off sequentially, but that requires remembering them

$ docker compose ps
NAME              IMAGE      COMMAND                  SERVICE    CREATED         STATUS                   PORTS
init-postgres-1   postgres   "docker-entrypoint.s…"   postgres   2 minutes ago   Up 2 minutes (healthy)   5432/tcp

# next, up --wait, could start the rest and it would ignore psql and init because they are behind a profile;
# it could even be done by the init itself if you can assume enough (e.g. about docker's sock)
$ docker compose up --wait
[+] Running 1/1
 ✔ Container init-postgres-1  Healthy

$ echo $?
0

# PS
$ docker compose --profile .init logs psql
psql-1  | You are connected to database "postgres" as user "postgres" on host "postgres" (address "172.28.0.2") at port "5432".

[gmile]

Just hit this too and confirmed this is still a problem as of 2.24.7. This thread doesn't seem to lack examples, but I'll include mine just in case too.

services:
  minio-create-buckets:
    image: quay.io/minio/mc
    environment:
      MINIO_HOST: http://minio:9000
      MINIO_ACCESS_KEY: minio-user
      MINIO_SECRET_KEY: minio-password
    entrypoint:
      - bash
      - -c
      - |
        mc alias set minio http://minio:9000 minio-user minio-password
        mc mb minio/bucket-1
        mc mb minio/bucket-2
        mc anonymous set public minio/bucket-1
        mc anonymous set public minio/bucket-2

        echo "Hello, world!" > sample.txt

        mc cp sample.txt minio/bucket-1
        mc cp sample.txt minio/bucket-2
    networks:
      - docker-network
    depends_on:
      minio:
        condition: service_healthy

  minio:
    image: quay.io/minio/minio
    environment:
      MINIO_ROOT_USER: minio-user
      MINIO_ROOT_PASSWORD: minio-password
    command: server /data --console-address ":9001"
    ports:
      - 9000:9000
      - 9001:9001
    networks:
      - docker-network
    healthcheck:
      test: ["CMD", "mc", "ping", "--exit", "local"]
      start_period: 10s
      start_interval: 1s
      interval: 3s
      timeout: 2s
      retries: 5

networks:
  docker-network:

[pkoutsovasilis]

FWIW I opened this PR and your example @gmile executes successfully with --wait-allow-exit

[pkoutsovasilis]

@ndeloof who could be an appropriate reviewer for the PR mentioned above, ty 🙂