I'm using s6 (http://skarnet.org/software/s6/) in order to supervise multiple processes.
I have couple services in my /etc/s6/ folder and one of them just need to call the start action from an init.d script.
As of now the script is well started but then it's trying to re-start it again and again.
Is there a way to avoid this ?
If your service is a one-shot, i.e. it needs to run once and die, but not be kept alive, then it's not a good candidate for supervision. Process supervision is only good for long-running processes, a.k.a. daemons.
The correct solution to your problem is to take your one-shot script out of the set of supervised services, and run it at some point in your initialization procedure without trying to have s6-svscan/s6-supervise manage it.
You can do this with s6, by including s6-svc -O /etc/s6/servicefolder at the beginning of your run script. This will tell to s6 to not start this service again.
If s6 is running in a docker container and you can't control how it's started, then you can't use s6-svc -O /etc/s6/my_service.
Assuming s6's services in the container are located at /etc/s6/, you can do this instead:
docker-compose.yml:
volumes:
- ./data:/data # a mapped or named volume
- ./run:/etc/s6/custom/run:ro
run:
#!/bin/sh
# ensure not run (successfully) before
if [ -f /data/custom-init-performed ]; then
echo 'INFO: custom init already performed'
s6-svc -D /etc/s6/custom # prevent s6 from restarting service
exit 0
fi
# ensure container healthy
# if you're using healthchecks then use whatever url is applicable here
# if not, use another approach to determine whether container is ready
until [ $(curl -sf http://localhost:3000/api/healthz | grep -E '^ "status": "pass",$' | wc -l) = 1 ]; do
echo 'WARN: container not healthy yet, will retry...' >&2
sleep 5
done
# do once-only init work
# ...
# prevent script's core logic from running again
touch /data/custom-init-performed
# prevent s6 from restarting service
s6-svc -D /etc/s6/custom
exit 0
