Demo project of using the Docker HEALTHCHECK functionality for a balenaCloud project.
It's a simple web application, which has a status endpoint, through which it can signal that that the service is unhealthy. The Dockerfile's HEALTHCHECK command will query that endpoint, and if the check fails, it will restart the service automatically:
HEALTHCHECK --start-period=5s --timeout=30s --interval=5s --retries=1 \
CMD curl --silent --fail localhost:8080/status
Here the --start-period, --timeout, --interval, and --retries values are set to pretty short so that the demo works quickly, in your application you need to set them as appropriate (see the default values on the linked HEALTHCHECK docs page above).
To use this, deploy the code on balenaCloud, enable the Public Device URL, and open the page, showing the service's status and provide tools to trigger an unhealthy state.
Pressing the Break Things button will predictably break things inside the application, which puts the service in a state that the healthcheck will catch:
After the healthcheck kicks in, the service will be restarted in a good state, and the page will show a working state again, and the uptime will show that it was just restarted.
If logged in to the device's Host OS, balena ps will show the state of the containers, and whether or not they are healthy:
Any data exposed via the HEALTHCHECK is now directly inspectable in the balenaCloud dashboard, specifically in the Device Diagnostics section. Any output from the HEALTHCHECK itself is passed back and added as a check, to allow users to expose their own data. Read more here:
