add "run" command

[stefanb2]

To quote from docker-compose CLI documentation:

Runs a one-time command against a service. For example, the following command starts the web service and runs bash as its command.

docker-compose run web bash

We use several different docker-compose YAML files in our projects, with various one-shot "helper commands" for different tasks a developer might need to perform, e.g. database setup.

Those tasks are performed by standard non-project specific images provided by our internal repository. I.e. the developer doesn't have to install anything or remember a complicated command line to perform those tasks. He simply types e.g.

docker-compose -f docker-compose.dev.yml run mysql-init

on any project to initialize the MySQL database to such a state that he can start working on the project.

The YAML file provides an easy way to supply the parameters, e.g.

mysql-init:
  env_file:
    - .dev.env
  image: <repo name>/<path>/mysql-init:5.7
  network_mode: host

Furthermore the file is under project version control.

Without the run command I have to translate the exisiting YAML file contents to the following command line:

podman run --rm -it --env-file .dev.env --network host <repo name>:5004/<path>/mysql-init:5.7

[muayyad-alsadi]

it's now much easier to add more commands using @cmd_run(podman_compose, cmd_name, cmd_desc)

see CONTRIBUTING.md

[muayyad-alsadi]

@stefanb2 there are some missing features like passing custom -l and -e and -v.

please test

$ python3 podman-compose.py -f examples/busybox/docker-compose.yaml run -h
usage: podman-compose.py run [-h] [-d] [--name NAME] [--entrypoint ENTRYPOINT]
                             [-e KEY=VAL] [-l KEY=VAL] [-u USER] [--no-deps]
                             [--rm] [-p PUBLISH] [--service-ports] [-v VOLUME]
                             [-T] [-w WORKDIR]
                             service ...

create a container similar to a service to run a one-off command

positional arguments:
  service               service name
  command               comman and its args

optional arguments:
  -h, --help            show this help message and exit
  -d, --detach          Detached mode: Run container in the background, print
                        new container name.
  --name NAME           Assign a name to the container
  --entrypoint ENTRYPOINT
                        Override the entrypoint of the image.
  -e KEY=VAL            Set an environment variable (can be used multiple
                        times)
  -l KEY=VAL, --label KEY=VAL
                        Add or override a label (can be used multiple times)
  -u USER, --user USER  Run as specified username or uid
  --no-deps             Don't start linked services
  --rm                  Remove container after run. Ignored in detached mode.
  -p PUBLISH, --publish PUBLISH
                        Publish a container's port(s) to the host (can be used
                        multiple times)
  --service-ports       Run command with the service's ports enabled and
                        mapped to the host.
  -v VOLUME, --volume VOLUME
                        Bind mount a volume (can be used multiple times)
  -T                    Disable pseudo-tty allocation. By default `podman-
                        compose run` allocates a TTY.
  -w WORKDIR, --workdir WORKDIR
                        Working directory inside the container

$ python3 podman-compose.py -f examples/busybox/docker-compose.yaml run --rm redis /bin/sh -c 'cat /etc/passwd'
# podman run --rm -i --name=busybox_redis_tmp60155 --pod=busybox ...
root:x:0:0:root:/root:/bin/ash
bin:x:1:1:bin:/bin:/sbin/nologin
...
redis:x:999:1000:Linux User,,,:/home/redis:/sbin/nologin
0

if it's not detached I make it interactive so you can run /bin/sh
please tell me if this behaviour is correct

[stefanb2]

Looks like you beat me to the punch. I started to work on an implementation yesterday. Unfortunately your changes seem to have broken the script completely.

If I take HEAD (7e0a036) I can't get podman-compose.py run XXX to work at all: the command parser sees XXX as the command instead of run. If I checkout HEAD^ (eb72a71) then it seems podman-compose.py run XXX does what it is supposed to do.

But commit eb72a71 broke podman-compose.py -t identity up completely, so I can't really test your changes:

  File ".../podman-compose.py", line 364, in mount_desc_to_args
    mount_desc = mount_dict_vol_to_bind(compose, fix_mount_dict(mount_desc, srv_name, cnt_name))
  File ".../podman-compose.py", line 334, in mount_dict_vol_to_bind
    proj_name = compose.proj_name
AttributeError: 'PodmanCompose' object has no attribute 'proj_name'

Looks like a typo and that line should instead read:

    proj_name = compose.project_name

[muayyad-alsadi]

Looks like a typo and that line should instead read:

yes, I've fixed that typo.

[mariushoch]

I got this working, but only with #43

[agriffis]

This works for me on master, but one note, it always tries to re-create the pod:

✸ docker-compose run app /bin/bash
podman pod create --name=search --share net -p 8000:8000
Error: unable to create pod: error adding pod to state: name search is in use: pod already exists
125
...snip...
root@6bed1a5d3f2d:/app# exit

[westurner]

Would this make exec easier to implement or would that need a separate implementation?

[stacyharper]

Got a prob with the run command

#!/bin/sh -x
#/entrypoint.sh

echo "$@"
exit 1

$ podman-compose run app foo bar
+ echo
+ exit 1

It seems my args are not passed to the entrypoint.

Is it a bug? Have I missed something?

[muayyad-alsadi]

I'll try to look at issue this weekend.