Docker Exit

[wandersonwhcr]

Context

This PR adds docs about how to implement HEALTHCHECK on Dockerfiles.

Also, adds a warning about HEALTHCHECK ignored by Kubernetes. I tried to search a great documentation on Internet about this behavior on K8S and reference it, but I didn't find something awesome.

Changes

• Tests included
• Documentation updated
• Commit message is clear

[wandersonwhcr]

@fernandesGabriel what do you think about this?

[wandersonwhcr]

Hey @s4mur4i what do you think about this?

[s4mur4i]

Hey @s4mur4i what do you think about this?

In our environment we use kubernetes based environment. We have been working with following variables:

          readinessProbe:
            exec:
              command:
                - php-fpm-healthcheck
            initialDelaySeconds: 5
            periodSeconds: 30
            timeoutSeconds: 10
            successThreshold: 1
            failureThreshold: 3
          livenessProbe:
            exec:
              command:
                - php-fpm-healthcheck
            initialDelaySeconds: 5
            periodSeconds: 30
            timeoutSeconds: 10
            successThreshold: 1
            failureThreshold: 3
            ```
I think values can be raised a little to not DDOS php-fpm-healthcheck.
What I could recommend maybe:

--interval=10s
--timeout=5s
--start-period=5s

In our case at higher load we see false negatives because of lower timeouts. that's why we raised it.

[wandersonwhcr]

But about Docker, do you think this doc is interesting?

[s4mur4i]

One thing confuses me:
https://github.com/renatomefi/php-fpm-healthcheck/blob/master/php-fpm-healthcheck#L21-L25
according to doc here, you can have exit 2 for example, which should not be used.

So shouldn't that be handled somehow? or we can have unforeseen behavior.
Since I am on primarily on kubernetes, I cant really say if interesting or not.

[wandersonwhcr]

Yes, this is confusing, but I think our script must not handle this.

Docker only accepts exits 0 to healthy, 1 to unhealthy and 2 that should not be used. php-fpm-healthcheck can return other values because there is more problems that can be found instead of binary 0 or 1.

To resolve this, I added to docs a || exit 1 after our script, described here:

https://github.com/renatomefi/php-fpm-healthcheck/pull/43/files#diff-b335630551682c19a781afebcf4d07bf978fb1f8ac04c6bf87428ed5106870f5R167-R168

Thinking about the problem, yes, we can add a --docker argument, but I think we don't need to this right now.

@s4mur4i I asked your help because nobody is replying my messages 🤣 sorry to disturb you

[fernandesGabriel]

@wandersonwhcr sorry for the late reply.

Unfortunately, I don't think I can opiate much here. @s4mur4i configurations are close to mine by the way.

Regarding the use of HEALTHCHECK on Kubernetes, looks like they even explicitly disabled it some time ago: kubernetes/kubernetes#50796. However, in the Kubernetes context, I would still prefer to use liveness/readiness configurations instead of something defined on the Dockerfile.

[wandersonwhcr]

@fernandesGabriel ty for you reply 😄

Yes, this change also adds a warning about Kubernetes ignoring HEALTHCHECK command from Docker.

BTW, TY ❤️