Closed IgnisDa closed 7 months ago
This just runs docker exec as shown here. Do we need to respecify the entrypoint when doing that?
Maybe. Because the following procfile works for me (which means entrypoint is respected for Procfiles)
release: migrate up # dont need to specify hanko
web: serve public
ENTRYPOINT for base image: https://github.com/teamhanko/hanko/blob/a4288afcf4426dc02e8f7825ea7c76454402028b/backend/Dockerfile#L41
I don't think docker exec
respects the entrypoint, though this is definitely what I'd expect, behavior-wise, from healthchecking.
How should we handle shell-based entrypoints? These don't allow command-line arguments at all, so healthchecks using them wouldn't work...
One interesting thing I've noted is that none of the herokuish containers have an entrypoint, which means that we're re-implementing that on our end in Dokku.
I think we can change that in Dokku for the builder-herokuish
builder by:
builder-release
to /start
docker container run
command in scheduler-run
to use --entrypoint /exec
, and dropping the EXEC_CMD
functionality/start
from the START_CMD
in scheduler-deploy-process-containerThe annoying thing is that we wouldn't be able to support &&
in release commands anymore - they are invoked differently in dokku - though that might actually be good from a parity perspective.
I think for this issue, we should use the /start
command for herokuish-based images, and otherwise respecting the entrypoint. I don't see a need to re-implement everything else.
I think for this issue, we should use the /start command for herokuish-based images, and otherwise respecting the entrypoint. I don't see a need to re-implement everything else.
Yep that sounds good.
However, I don't really use herokuish based images so I can't comment on those.
Can you give me a valid config.yaml for this hanko test? The container doesn't run as is, even with passing args:
% docker run -it --rm hanko serve all
2023/10/01 06:22:05 Using config file: ./config/config.yaml
2023/10/01 06:22:05 failed to validate config: failed to validate passcode settings: failed to validate smtp settings: smtp host must not be empty
Here is a docker compose dev container: https://github.com/IgnisDa/codefarem/blob/main/.devcontainer/docker-compose.yml
Description of problem
As the title says.
Steps to reproduce
docker-container-healthchecker version
0.6.4
Output of failing command
No response