Claus Ibsen created CAMEL-25088:
-----------------------------------
Summary: camel-health - Health checks: fix bugs found in a deep
review
Key: CAMEL-25088
URL: https://issues.apache.org/jira/browse/CAMEL-25088
Project: Camel
Issue Type: Bug
Components: camel-core
Reporter: Claus Ibsen
A review of the health checks in camel-health (and camel-main) found these bugs:
# With the {{default}} and {{oneline}} exposure level the result only includes
the DOWN checks (or else the first check), so a check in the UNKNOWN state
(such as a disabled check, a consumer with
{{healthCheckConsumerEnabled=false}}, or {{camel.health.initialState=UNKNOWN}})
is left out and Camel reports ready, while the {{full}} exposure level reports
not ready for the same state.
# {{camel.health.enabled=false}} does not disable the health check registry
(the early return in BaseMainSupport skips setting enabled=false), so the
checks are still loaded and invoked.
# RoutesHealthCheckRepository and ConsumersHealthCheckRepository never remove
the check of a removed route, so each removed route (route reload, routes from
templates/kamelets) is kept in memory.
# A health check that throws an exception fails the whole health check (the
endpoint returns an error with no details) instead of the check being reported
as DOWN.
# RouteHealthCheck throws NullPointerException when the route has been removed
(such as during route reload).
# DefaultHealthCheckRegistry.register rejects a check whose id only resembles
another check (a custom check {{foo}} is rejected when a route {{foo}} exists,
{{x}} is rejected when {{x-health-check}} is registered).
# RouteHealthCheck reports the initial state of the registry for a suspended
route (UP when initialState=UP), instead of DOWN as for a stopped route.
# The exclude pattern does not allow spaces around the comma ({{foo, bar}} does
not exclude {{bar}}).
Not changed:
* The result strategy (HealthCheckResultStrategy) runs after the
success/failure counters are updated.
* ScheduledPollConsumer.lastErrorDetails is a plain HashMap that is read by the
health check thread.
* The health dev console lists the context check twice (readiness and liveness).
_Claude Code on behalf of Claus Ibsen_
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
