How to add a custom health check#

This guide shows you how to add your own liveness or readiness check. Use a liveness check for something that proves the process itself is alive, and a readiness check for a dependency the process needs in order to serve traffic.

For the interface members, see Interfaces.

Add a liveness check#

A liveness check must not access ZODB and must not block. Implement ILivenessCheck and register it as a named utility.

from zope.interface import implementer
from plone.observability.interfaces import ILivenessCheck


@implementer(ILivenessCheck)
class MyLivenessCheck:
    name = "myapp"

    def __call__(self):
        return True, "all good"

<utility
    factory=".checks.MyLivenessCheck"
    provides="plone.observability.interfaces.ILivenessCheck"
    name="myapp"
    />

Add a readiness check#

A readiness check may access ZODB and any dependency the process needs to serve requests. Implement IReadinessCheck.

from zope.interface import implementer
from plone.observability.interfaces import IReadinessCheck


@implementer(IReadinessCheck)
class MyReadinessCheck:
    name = "myapp"

    def __call__(self):
        ok = _check_external_service()
        return ok, "service ok" if ok else "service unavailable"

<utility
    factory=".checks.MyReadinessCheck"
    provides="plone.observability.interfaces.IReadinessCheck"
    name="myapp"
    />

The check name appears under checks in the probe response, keyed by the name attribute.

Readiness checks run together, one run at a time, and the result is reused for a short time. A check does not need its own locking or caching against concurrent probes. It should still return within a few seconds: while a run is stuck, the pod reports not ready. See About health probes.