Operator crashes at startup with NPE on MicrometerMetrics.reconcileCustomResource

[sclorng]

Bug Report

What did you do?

I start the operator.

What did you expect to see?

Operator running

What did you see instead? Under which circumstances?

2022-01-15 17:48:30,163 ERROR traceId= [io.qua.run.Application] (main) Failed to start application (with profile prod): java.lang.NullPointerException
at io.javaoperatorsdk.operator.monitoring.micrometer.MicrometerMetrics.reconcileCustomResource(MicrometerMetrics.java:63)
at io.javaoperatorsdk.operator.processing.event.EventProcessor.submitReconciliationExecution(EventProcessor.java:150)
at io.javaoperatorsdk.operator.processing.event.EventProcessor.handleMarkedEventForResource(EventProcessor.java:134)
at io.javaoperatorsdk.operator.processing.event.EventProcessor.handleAlreadyMarkedEvents(EventProcessor.java:364)
at io.javaoperatorsdk.operator.processing.event.EventProcessor.start(EventProcessor.java:356)
at io.javaoperatorsdk.operator.processing.event.EventSourceManager.start(EventSourceManager.java:77)
at io.javaoperatorsdk.operator.processing.Controller.start(Controller.java:197)

Environment

Kubernetes cluster type:

vanilla

$ Mention java-operator-sdk version from pom.xml file

2.0.0

$ java -version

native GraalVM 21.3

$ kubectl version

Client Version: version.Info{Major:"1", Minor:"18", GitVersion:"v1.18.8", GitCommit:"9f2892aab98fe339f3bd70e3c470144299398ace", GitTreeState:"clean", BuildDate:"2020-08-13T16:12:48Z", GoVersion:"go1.13.15", Compiler:"gc", Platform:"linux/amd64"}
Server Version: version.Info{Major:"1", Minor:"18", GitVersion:"v1.18.6", GitCommit:"dff82dc0de47299ab66c83c626e08b245ab19037", GitTreeState:"clean", BuildDate:"2020-07-15T16:51:04Z", GoVersion:"go1.13.9", Compiler:"gc", Platform:"linux/amd64"}

Possible Solution

Additional context

I'm upgrading several operators and this is the first one that failed at startup. It has two reconciler based on the same abstract class. Using Quarkus Sdk Operator 3.0.0. I import the quarkus-micrometer extension for my own use. Both controller has no event source.

[sclorng]

Looking at the code, seems retryState for the marked event source ResourceID don't exist. Don't know if it's normal at startup. Nevertheless maybe MicrometerMetrics.reconcileCustomResource should not assume retryInfo couldn't be null.

[scrocquesel]

EventProcessorTest.startProcessedMarkedEventReceivedBefore would produce the NPE if micrometers implementation was used instead of Metrics.NOOP

I was proposing this fix in micrometer implementation

public void reconcileCustomResource(ResourceID resourceID,
      RetryInfo retryInfoNullable) {
    Optional<RetryInfo> retryInfo = Optional.ofNullable(retryInfoNullable);
    incrementCounter(resourceID, RECONCILIATIONS + "started",
        RECONCILIATIONS + "retries.number", "" + retryInfo.map(x -> x.getAttemptCount()).orElse(0),
        RECONCILIATIONS + "retries.last", "" + retryInfo.map(x -> x.isLastAttempt()).orElse(true));
  }

but I think that changing signature for Optional<RetryInfo> is worthwhile. Still it would be a breaking change.

On testing, there is no test on the micrometer implementation, so I opt for verifying mock call with null value.
But it seems there is no test for verifying call to metrics methods. EventProcess ctors does not allow to pass a Metrics object. Should I add an overload ctor or mock the entire

eventSourceManager
                .getController()
                .getConfiguration()
                .getConfigurationService()
                .getMetrics()

chain call ?

[csviri]

thank you @sclorng @scrocquesel. Will close this now if not objections.