JarCastingAsync Metrics Using Guice AOP
Leverage Codahale metrics (or any Metric Registry for that matter) to provide metrics on methods that complete asynchronously. The library requires Guice AOP to work (bundled by default)
Maven Artifacts
This project is available on Maven Central. To add it to your project you can add the following dependency to your pom.xml:
<dependency>
<groupId>com.github.isopropylcyanide</groupId>
<artifactId>async-codahale-metrics</artifactId>
<version>1.0</version>
</dependency>
Features
| Annotation | Codahale Equivalent |
|---|---|
| @AsyncTimed | @Timed |
| @AsyncMetered | @Meter |
| @AsyncExceptionMetered | @ExceptionMetered |
Usage
- Initialise the Module in your Guice Injector
...
MetricRegistry registry = new MetricRegistry();
Guice.createInjector(new AsyncMetricsModule(metricRegistry));
...
- Annotate the required methods with the required annotation
@AsyncTimed
@AsyncExceptionMetered
CompletableFuture<Integer> asyncMethodThatCompletesNormally() {
return CompletableFuture
.supplyAsync(() -> StringUtils.split(getClass().getName(), "."))
.thenApply(s -> s.length);
}
@AsyncMetered
CompletableFuture<String> asyncMethodThatCompletesNormally() {
return CompletableFuture.supplyAsync(() -> "Hello World");
}
Why another metrics library?
We wish to leverage Dropwizard metrics for async callbacks. The likes of @Timed or @ExceptionMetered for methods that return Futures or CompletableFutures or Callbacks
These annotation for marking a method of an annotated object make the code much more readable and separate the unnecessary boilerplate of marking metrics from the business logic
@Timed @Timed
@ExceptionMetered @ExceptionMetered
public X getX() { public Future<X> getX() {
//business logic //business logic that might execute asynchronously
return x; return Future.of(x)
} }
Works as expected Doesn't work as expected
However, the second approach won't produce correct results as if the calling thread dispatches the work to another thread in the pool then this method execution completes without waiting for the result. Ideally, we want to mark our metrics and figure out a way to do meta stuff once the callback resolves either successfully or exceptionally. That is why this library as created.
Enhancements
- Feel free to extend this to mark any
custom metrics. All you need is an annotation and the corresponding aspects - Instead of
Codahale Metric Registryit can be extended to any registry. - Instead of
completable future, it can be any random callback. The only requirement is of a hook to execute action post callback resolution that doesn't block.
Testing
- Create a driver class with a new metric registry
- Install the module
- Set up a local Console reporter
ConsoleReporter reporter = ConsoleReporter.forRegistry(metrics)
.convertRatesTo(TimeUnit.SECONDS)
.convertDurationsTo(TimeUnit.MILLISECONDS)
.build();
reporter.start(1, TimeUnit.SECONDS);
Wiki
https://medium.com/@aman_garg/leveraging-async-metrics-using-aspects-81838b9b887e
Support
Please file bug reports and feature requests in GitHub issues.
License
Copyright (c) 2012-2020 Aman Garg
This library is licensed under the Apache License, Version 2.0.
See http://www.apache.org/licenses/LICENSE-2.0.html or the LICENSE file in this repository for the full license text.