Stopping docker container from inside

Question

I have a cronjob running inside a docker container that checks whether all services are running as expected. If this cronjob determines that there is a problem I would like to stop the docker container (from inside...)

Unfortunately exit just stops my cronjob script

Answer 1

Basically, you need PID 1 to exit to stop the container.

Doing kill -s SIGKILL 1 won't work because PID 1 is protected.

As suggested by @Thomasleveil, you could add code such as trap "exit" SIGINT SIGTERM to the PID 1 script, which will mean the process will exit when sent a kill -s SIGINT 1. I slightly prefer this method to the one you came up with (killing the child process directly) as it gives the parent process a chance to clean up and also the parent process should be able to find the PID of the child process without awk.

The easiest way to handle this is with some sort of supervisor process, and handily Docker now provide one called "tini". Just run add the argument --init to docker run and Docker will set up tini as PID 1 in the container. A full description is here: https://docs.docker.com/engine/reference/run/#specify-an-init-process

Answer 2

I faced the same problem and I solved it by using healthcheck feature of docker.

You can add checks to verify your services in a file like heahthcheck.sh. Then append something like following to your Dockerfile :

# Healthcheck
COPY healthcheck.sh /
RUN chmod +x /healthcheck.sh

HEALTHCHECK --interval=10s --retries=5 CMD /healthcheck.sh

This makes sure that docker checks the health of your container by running your script in specified interval and if it fails 5 times in a row, it marks the state unhealthy.

Now, in order to restart your container if the state goes unhealthy you can use autoheal like this:

docker run -d \
    --name autoheal \
    --restart=always \
    -e AUTOHEAL_CONTAINER_LABEL=all \
    -v /var/run/docker.sock:/var/run/docker.sock \
    willfarrell/autoheal

This will restart your container every time its state goes unhealthy.

Answer 3

Inspired by @JakeRobb: You can execute the docker with the command:

/bin/bash -c "while true; do sleep infinity || exit 0; done"

This will execute sleep on a PID lower than 1. Once you kill it, then the loop will exit. For that, from within the docker use:

pkill -f sleep

Answer 4

I had to figure this out recently. After much frustration, what I landed on is this:

In your dockerfile, specify an ENTRYPOINT:

ENTRYPOINT ["/entrypoint.sh"]

Then, provide such a script. It doesn't need to do anything other than invoke your application. Feel free to add additional setup to the script, but be mindful that if the script does anything after it invokes your application, it could mask your application's return code. If that's relevant to you, make sure to have the script capture the return code and propagate it out as its own return code.

Note: Do not use exec to invoke your application!

The result will be that PID 1 belongs to entrypoint.sh, and your application will have some other PID. Mine tend to land on PID 9 or 10.

Then, when you need to kill your application, simply determine its PID and invoke kill -SIGKILL $PID, or pkill -SIGKILL yourapp. Your process, not being PID 1, will receive the signal and exit immediately. Your entrypoint.sh will promptly exit, because that's what you designed it to do after your process exits.

Answer 5

My example of a self-killed /usr/local/bin/docker-entrypoint.sh

#!/bin/sh
#

# Following guard prevents to stuck containers in DIND GitLab Runner
# If container still works in 300 seconds, this will kill yourself
THE_PID=$$
if [ $THE_PID -eq 1 ]; then
    # Docker does not allow kill process 1, so restart as child process
    /usr/local/bin/docker-entrypoint.sh
    exit $?
fi
(sleep 300; echo "Killing process: $THE_PID"; kill -9 $THE_PID) &


YOUR_SCRIPT_TEXT....

Answer 6

I kill the "blocking process" i.e. the main process you docker is running; like so (I this case docker is running a sleep infinity command for demonstration purposes).

ps -afx | grep sleep | awk '{print $1}' | xargs kill -9

Answer 7

I tried to kill process 1 without success.

Give a try to the comment of @zero323 with shutdown -h now. It works fine (sorry, I can not vote for it directly since it is not in the list of answers).