Intermittent TestServer.checkMetricsForExecutorService

[romain-grecourt]

java.lang.AssertionError: 
Completed task count after accessing slow endpoint
Expected: is <1>
     but: was <0>
	at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:20)
	at io.helidon.metrics.TestServer.checkMetricsForExecutorService(TestServer.java:221)
	at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104)
	at java.base/java.lang.reflect.Method.invoke(Method.java:578)
	at org.junit.platform.commons.util.ReflectionUtils.invokeMethod(ReflectionUtils.java:688)

[tomas-langer]

It seems there is an inherent race condition in this test - the update of metrics may happen after the response is returned, causing the next request to be "too early" to see the modified metric.
This should be refactored to use some deterministic appraoch

[tomas-langer]

For main branch, we will disable this test until a solution is found

[tjquinno]

Yes, this is the case in at least one other test as well. I am addressing them.

[danielkec]

Just failed in 4x on top of the fix commit ce733cb

[INFO] Running io.helidon.metrics.TestServer
Error:  Tests run: 8, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 1.559 s <<< FAILURE! - in io.helidon.metrics.TestServer
Error:  io.helidon.metrics.TestServer.checkMetricsForExecutorService  Time elapsed: 1.021 s  <<< FAILURE!
java.lang.AssertionError: 
Completed task count after accessing slow endpoint
Expected: is <1>
     but: was <0>
	at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:20)
	at io.helidon.metrics.TestServer.lambda$checkMetricsForExecutorService$1(TestServer.java:223)
	at io.helidon.common.testing.junit5.MatcherWithRetry.retry(MatcherWithRetry.java:76)
	at io.helidon.metrics.TestServer.checkMetricsForExecutorService(TestServer.java:207)
	at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104)

[danielkec]

https://github.com/helidon-io/helidon/pull/5047/files#r987852724

[danielkec]

Tweak with FT #5103

[tjquinno]

Re-opening. Daniel had improved on my original retry approach, but retrying is not the right solution here. The test GreetService already has a CountdownLatch which other tests use but this one did not. I'll submit yet another PR that fixes the test so it correctly uses the latch.