Metrics client for Java

An application metrics client for Java integrated with Graphite/OpenTSDB

License

License

Categories

Categories

Metrics Application Testing & Monitoring Monitoring Prometheus
GroupId

GroupId

com.adobe.aam
ArtifactId

ArtifactId

metrics-sample-prometheus
Last Version

Last Version

1.0.3
Release Date

Release Date

Type

Type

jar
Description

Description

Metrics client for Java
An application metrics client for Java integrated with Graphite/OpenTSDB
Project URL

Project URL

https://github.com/adobe/metrics-client-for-java
Source Code Management

Source Code Management

https://github.com/adobe/metrics-client-for-java.git

Download metrics-sample-prometheus

How to add to project

<!-- https://jarcasting.com/artifacts/com.adobe.aam/metrics-sample-prometheus/ -->
<dependency>
    <groupId>com.adobe.aam</groupId>
    <artifactId>metrics-sample-prometheus</artifactId>
    <version>1.0.3</version>
</dependency>
// https://jarcasting.com/artifacts/com.adobe.aam/metrics-sample-prometheus/
implementation 'com.adobe.aam:metrics-sample-prometheus:1.0.3'
// https://jarcasting.com/artifacts/com.adobe.aam/metrics-sample-prometheus/
implementation ("com.adobe.aam:metrics-sample-prometheus:1.0.3")
'com.adobe.aam:metrics-sample-prometheus:jar:1.0.3'
<dependency org="com.adobe.aam" name="metrics-sample-prometheus" rev="1.0.3">
  <artifact name="metrics-sample-prometheus" type="jar" />
</dependency>
@Grapes(
@Grab(group='com.adobe.aam', module='metrics-sample-prometheus', version='1.0.3')
)
libraryDependencies += "com.adobe.aam" % "metrics-sample-prometheus" % "1.0.3"
[com.adobe.aam/metrics-sample-prometheus "1.0.3"]

Dependencies

compile (11)

Group / Artifact Type Version
com.typesafe : config jar 1.3.2
com.google.guava : guava jar 16.0
org.apache.commons : commons-lang3 jar 3.1
net.jodah : failsafe jar 1.0.3
com.fasterxml.jackson.core : jackson-databind jar 2.9.4
com.adobe.aam : metrics-guice jar 1.0.3
org.eclipse.jetty : jetty-servlet jar 9.4.11.v20180605
io.prometheus : simpleclient_hotspot jar 0.4.0
io.prometheus : simpleclient_servlet jar 0.4.0
org.slf4j : slf4j-api jar 1.7.25
ch.qos.logback : logback-classic jar 1.2.3

test (6)

Group / Artifact Type Version
junit : junit jar 4.11
org.hamcrest : hamcrest-all jar 1.3
org.spockframework : spock-core jar 1.1-groovy-2.4
org.mockito : mockito-core jar 1.9.5
org.powermock : powermock-mockito-release-full jar 1.6.4
net.bytebuddy : byte-buddy jar 1.8.11

Project Modules

There are no modules declared in this project.

Metrics Library for Java

Build Status

An application metrics client integrated with Graphite/OpenTSDB/more to come. Offers the following features:

  • sends the metrics in batch to the backend
  • fail safe (retry + circuit breaker to not overwhelm the backend)
  • send to multiple backends in parallel
    • useful for migrating to one backend to another
    • useful for sending a set of metrics to one backend (e.g. a sub-set of importants metrics used for alerts in Prometheus) and another set to another backend (e.g. all metrics to OpenTSDB).

Usage

Maven

<!-- https://mvnrepository.com/artifact/com.adobe.aam/metrics-all -->
<dependency>
    <groupId>com.adobe.aam</groupId>
    <artifactId>metrics-all</artifactId>
    <version>1.0.2</version>
</dependency>

Gradle

// https://mvnrepository.com/artifact/com.adobe.aam/metrics-all
compile group: 'com.adobe.aam', name: 'metrics-all', version: '1.0.2'

Sample application

You can find a demo app in this project: Sample Application

Java docs

You can find here the Java docs

Architecture

EC2 Shredder diagram

Sample App: How to use the metrics library

See metrics-sample app, for a fully fledged demo.

Create a metric

There are several metric types available: CounterMetric, AverageMetric, MaxMetric, MinMetric etc.

Average Metric

// Create a metric that generates the average of the values.
Metric metric = Metric.newInstance("request.time", Metric.Type.AVG);

metric.track(10);
metric.track(20);

metric.get(); // returns (10 + 20) / 2 = 15

Counter

CounterMetric metric = new CounterMetric("event.pixel");

metric.increment(); // New value is 1
metric.increment(); // New value is 2
metric.add(10); // New value is 12

metric.get(); // returns 12

Send a metric to a backend (e.g. Graphite / OpenTSDB)

How the metric client manages to publish metrics to the backend

By default, the client is able to send metrics to Backend using a retry mechanism guarded by a circuit breaker mechanism. The former ensures metrics are being resent should the initial request fail, providing a way of not losing important metrics. On the other hand, the circuit breaker mechanism makes the client silently aware of a non-responsive Backend backend by discarding all incoming metrics until the service is up and running again.

Nevertheless, one can always switch to a client w/o retry and/or w/o circuit breaker mechanisms depending on the environment needs. For instance, for a testing environment that does not highly rely on metrics, the metric client can be used without retry and circuit breaker mechanisms.

However, there is a great benefit brought by the enablement of both mechanism (via config) - safely and effectively (using retry) sending metrics while not polluting with metrics the memory of the metric client when the Backend is non-responsive. The downside of enabling the circuit breaker is the potential impact on aggregated metrics - each client node might enter into an open circuit at a different time, hence the metrics are prone to inconsistency

References:

Create a metric client

The MetricClientFactory contains a series of methods to create a metric client. It can use either configuration file(s), a Properties object, a typesafe config etc.

monitor {
    collectFrequency : 60000ms

    tags {
        env : prod
        app_name: mywebapp
        region: us-east-1
        cluster: edge1
        useHostname: true
    }

    publishers: [ 
        {
            name: Graphite Primary
            type: graphite
            host: graphiterelay.com
            port: 2003
            batch_size: 500
            sendOnlyRecentlyUpdatedMetrics: true
            resetCounters: true
            filter.allowList : [
              // Only these metrics will be sent through this client.
              "*"
            ]
        },
        {
          name: Prometheus
          type: prometheus
          sendOnlyRecentlyUpdatedMetrics: false
          resetCounters: false
          filter.allowList : [
            // Only these metrics will be exposed through this client.
            "*"
          ]
          relabel: {
            // Relabel example. Useful for 3rd party metrics such as those coming from Codahale.
            // myapp.db.table1.requests -> myapp_db_requests{db_table="table1"}
            "db\\.([^.]+).*": [
              {
                "db_table": "$1"
              }
            ]
          }
        }
    ]
}
// Create metric client.
MetricClient metricClient = new MetricClientFactory()
      .create(config.get("monitor.publishers"));

Send metrics

long timeNow = System.currentTimeMillis() / 1000;
metricClient.sendAndReset(metric, timeNow);
metricClient.flush();

Using a metric agent to monitor your metrics

The metric agent monitors a list of metrics. Every 60 seconds (or other configurable frequency), it sends the current metric values to the specified metric client and resets them to zero.

List<Metric> metrics = new ArrayList<>();
// Add metrics to the list
...

MetricAgentConfig config = ImmutableMetricAgentConfig.builder()
                .collectFrequency(config.getDuration("monitor.collectFrequency"))
                .addMetrics(metrics)
                .tags(MetricAgentConfig.tagsFromConfig(config.getConfig("monitor.tags")))
                .build();

MetricAgent metricAgent = new MetricAgent(metricClient, config);

metricAgent.startAsync();

...

metricAgent.stopAsync();

Codahale integration

This metrics library is integrated with codahale. You can add one or more codahale MetricRegistry to the metric agent and they will be reported to the backend. Advantages for doing this:

  • sends the metrics in batch to the backend
  • fail safe (retry + circuit breaker)
  • resets the metrics once sent to the backend (useful for Codahale Counters due to https://github.com/dropwizard/metrics/issues/143)
  • send to multiple backends in parallel

Build

To build this project:

$ git clone [email protected]:adobe/metrics-client-for-java.git
$ cd metrics-client-for-java/
$ ./gradlew build

Bugs and Feedback

For bugs, questions and discussions please use the GitHub Issues.

LICENSE

Copyright 2018 Adobe Systems Incorporated

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

com.adobe.aam

Adobe, Inc.

Open source from Adobe

Versions

Version
1.0.3
1.0.2