Counters not idempotent / state & lifecycle unclear

[dylanraithel]

Counters (and presumably other metering classes) don't appear to be idempotent. A single instance of a Counter class, registered to a DataDogRegistry doesn't produce consistent results when calling count() or measure(). The below example is from a scala sbt console and is a verbatim output, in order.

res74: io.micrometer.core.instrument.Counter = io.micrometer.core.instrument.step.StepCounter@34cb4a37

scala> counter.count()
res75: Double = 0.0

scala> counter.increment(10)

scala> counter.count()
res77: Double = 0.0

scala> counter.measure()
res78: Iterable[io.micrometer.core.instrument.Measurement] = [Measurement{statistic='COUNT', value=0.0}]

scala> counter.increment(10)

scala> counter.measure()
res80: Iterable[io.micrometer.core.instrument.Measurement] = [Measurement{statistic='COUNT', value=20.0}]

scala> counter.count()
res81: Double = 20.0

scala> counter.count()
res82: Double = 0.0

What's the lifecycle of the state of this object supposed to look like? Should the count values ever reset on their own? I'd really like to use this library and not go back to Dropwizard metrics if I don't have to

[dylanraithel]

Here's another example:

scala> counter.increment(10)

scala> counter.increment(10)

scala> counter.increment(10)

scala> counter.increment(10)

scala> counter.increment(10)

scala> counter.increment(10)

scala> counter.increment(10)

scala> counter.increment(10)

scala> counter.count()
res107: Double = 0.0

scala> counter.measure()
res108: Iterable[io.micrometer.core.instrument.Measurement] = [Measurement{statistic='COUNT', value=0.0}]

scala> counter.measure()
res109: Iterable[io.micrometer.core.instrument.Measurement] = [Measurement{statistic='COUNT', value=0.0}]

scala> counter.count()
res110: Double = 0.0

scala> counter.increment()

scala> counter.measure()
res112: Iterable[io.micrometer.core.instrument.Measurement] = [Measurement{statistic='COUNT', value=101.0}]

scala> counter.count()
res113: Double = 101.0```

[dylanraithel]

It seems the call to a parameter-less increment method resolves the missing counted values with the state of the object. This doesn't make much sense to me :D

[checketts]

That is bizarre behavior and not at all what I would expect (especially based on my understanding of how it is implemented).

Increment takes a double. Is it possible the Scala 10 isn't calling the method correctly? How does it behave if you use 10.0?

[dylanraithel]

It behaves the same, but I do eventually see the metrics in DataDog, so are the values in the Counter instance(s) just being reset after the library purges data to DataDog?

res156: Double = 0.0

scala> counter.increment(10.0)

scala> counter.count
res158: Double = 0.0

scala> counter.measure()
res159: Iterable[io.micrometer.core.instrument.Measurement] = [Measurement{statistic='COUNT', value=10.0}]

scala> counter.count
res160: Double = 10.0

[shakuzen]

This, I believe, is due to timing and because the Step* implementations will return the value (count) for the previous interval. I'll open a separate issue to add some unit tests that demonstrate this for Step* metrics, as it has confused others and it would be nice to have something to point people to that also verifies the behavior. For now you can check the TCK test for counter increments.

micrometer/micrometer-core/src/main/java/io/micrometer/core/instrument/step/StepDouble.java

Lines 58 to 64 in 49162a8

	/**
	* @return The value for the last completed interval.
	*/
	public double poll() {
	rollCount(clock.wallTime());
	return previous;
	}

micrometer/micrometer-test/src/main/java/io/micrometer/core/tck/CounterTest.java

Lines 36 to 49 in 49162a8

	@DisplayName("multiple increments are maintained")
	@Test
	default void increment(MeterRegistry registry) {
	Counter c = registry.counter("myCounter");
	c.increment();
	clock(registry).add(step());
	assertThat(c.count()).isEqualTo(1.0, offset(1e-12));
	c.increment();
	c.increment();
	clock(registry).add(step());
	// in the case of a step aggregating system will be 2, otherwise 3
	assertThat(c.count()).isGreaterThanOrEqualTo(2.0);
	}

[github-actions]

If you would like us to look at this issue, please provide the requested information. If the information is not provided within the next 7 days this issue will be closed.

[github-actions]

Closing due to lack of requested feedback. If you would like us to look at this issue, please provide the requested information and we will re-open.