Unable to use docker because of the plugin default CMD

[SylvestreMassalaz]

Hi there, I currently have a problem when using this image that seems to come from the docker plugin configuration.

Either I leave the CMD field unmodified and the docker command fails because the startup script is not launched or I try to set it to the startup script and the SSH connection fails.

I didn't find any hints in the plugin documentation, but I noticed that the default command doesn't looks like the default one (see the result i got from a docker inspect below):

• The default one

"Path": "bash",
"Args": [
    "-c",
      "/usr/sbin/sshd -D -p 22"
],

• The custom one

"Path": "/jenkins-slave-startup.sh",
"Args": [],

Does anyone know how to set it correctly?

[mattrobenolt]

We do this:

We have an image that uses this as it's base, so something like:

FROM tehranian/dind-jenkins-slave

COPY docker-entrypoint.sh /entrypoint
ENTRYPOINT ["/entrypoint"]

Then our docker-entrypoint.sh:

#!/bin/bash
set -e

case "$1" in
  image | test | deploy)
    cd /home/jenkins && exec wrapdocker sudo -EHsu jenkins ci-${1}.sh;
esac

exec wrapdocker "$@"

Basically, everything just needs to call wrapdocker which is what controls starting the daemon in the background.

Does that make sense?

[tehranian]

@mattrobenolt Is changing our Dockerfile to work better out of the box an option?

[mattrobenolt]

Likely, I'm not even sure why sshd is needed tbh since we don't use it.

But I'd rather let @branden answer better since he mostly works on this stuff. I'm just a spectator mostly. :)

[SylvestreMassalaz]

Hi, thanks for your response!

Sadly, I'm still unable to make it work correctly, this time even with the default command set by the jenkins docker plugin.

As you can see I only added a few things when starting from your example

FROM tehranian/dind-jenkins-slave
MAINTAINER Sylvestre Massalaz <[email protected]>

ENV YML_DIR /etc/pl/

RUN curl -L -k https://github.com/docker/compose/releases/download/1.3.3/docker-compose-`uname -s`-`uname -m` > /usr/local/bin/docker-compose &&\
    chmod +x /usr/local/bin/docker-compose

RUN apt-get update &&\
    apt-get install -y \
      wget \
      curl \
      unzip

COPY pl-deploy-test/* ${YML_DIR}

COPY compose-entrypoint.sh /entrypoint

RUN chmod +x /entrypoint

ENTRYPOINT ["/entrypoint"]

My entrypoint is also really simple

#! /bin/bash

set -e

exec wrapdocker "$@"

When it's launched with the plugin default command, I get this with docker inspect

    "Path": "/entrypoint",
    "Args": [
        "bash",
        "-c",
        "/usr/sbin/sshd -D -p 22"

Then I tried to remove the bash (the command being already inside an exec) and I mannually set it to /usr/sbin/sshd -D, the ni got

    "Path": "/entrypoint",
    "Args": [
        "/usr/sbin/sshd",
        "-D"
    ],

Either way, the container is created by Jenkins but the SSH connexion seems to fail because it shows offline and never starts the build.

The only setting I made on the template configuration is checking "run privileged" as asked in the readme.

Any Idea?

Edit : I think that the error with ssh has something to do with the fact that wrapdocker seems to take a while, maybe never launching the default command. Maybe we should use a smaller version that only does the basic configuration and daemon starting?