docker-compose down does not cascade down SIGTERM through two layers of bash scripts

Question

I have two bash scripts, one as an entrypoint to my docker container, looking like this:

#!/bin/bash

sig_handler() 
{
  echo "[LAYER1] killing children with pid "$pid
  [[ $pid ]] && kill $pid
  exit 1
}

trap 'sig_handler' SIGINT SIGTERM

while true; do
  ./nextlayer.sh & pid=$!
  wait $pid
  echo "Waiting 5 seconds before starting a new worker..."
  sleep 5
done

Now nextlayer.sh has some kind of signal trapping too, BUT also tries to clean up some stuff, like so:

#!/bin/bash

sig_handler() 
{
  echo "[LAYER2] Exiting main script and cleaning up tasks"
  cleanup
}

cleanup() 
{
  echo "[LAYER2] Cleaning up"
  sleep 5
  echo "Cleanup done, exiting with SIGTERM"
  exit 143
}

trap 'sig_handler' SIGINT SIGTERM

i=0
while [ $i -lt 10 ]
do
  i=$(( $i + 1 ))
  sleep 1
  echo $i
done

This just loops over i while i is < 10 and echoes the result every second. When running the entrypoint script locally without docker and pressing ctrl + c to exit, the script terminates as expected and outputs the cleanup function content.

However, when applying the same logic onto a docker container, the SIGTERM only reaches the sighandler in the entrypoint script, but never in the second layer script. Anyone know what I'm doing wrong?

Answer 1

I found the solution.

in the sig_handler function of the entrypoint script, the following wait instruction solves it:

sig_handler() 
{
    echo "[LAYER1] killing children with pid "$pid
    [[ $pid ]] && kill $pid
    wait $pid # this is crucial
    exit 1
}

So before actually quitting the container, the wait $pid forces it to actually wait out the exit from the subsequent script. I tested this 5 iterations of scripts, it all works.