Skip to content

Latest commit

 

History

History
362 lines (268 loc) · 14.2 KB

README.md

File metadata and controls

362 lines (268 loc) · 14.2 KB
Raw

Statful Client for Java

Build Status

Statful client for Java. This client is intended to gather metrics and send them to Statful.

Table of Contents

Supported Versions of Java

Statful client Version Tested Java versions
2.x.x Java 8

Configuration

Add one of the following snippets to your POM.xml file.

Configuration for UDP client

<dependency>
    <groupId>com.statful.client</groupId>
    <artifactId>udp-client</artifactId>
    <version>${statful-client.version}</version>
</dependency>

Configuration for HTTP client

<dependency>
    <groupId>com.statful.client</groupId>
    <artifactId>http-client</artifactId>
    <version>${statful-client.version}</version>
</dependency>

Using AspectJ

Configuration

Configure your project to weave your application as you like but don't forget to include the following dependencies on your project:

<dependency>
    <groupId>com.statful.client</groupId>
    <artifactId>aspects</artifactId>
    <version>${statful-client.version}</version>
</dependency>

<dependency>
    <groupId>org.aspectj</groupId>
    <artifactId>aspectjrt</artifactId>
    <version>${aspectj.version}</version>
</dependency>

Then, you must set StatfulAspect with your StatfulClient instance:

StatfulAspect statfulAspect = Aspects.aspectOf(StatfulAspect.class);
statfulAspect.setStatfulClient(statfulClient());

You must include the aspect on your AspectJ configuration:

<weaver>
    <include within="com.statful.client.aspects.*"/>
</weaver>
<aspects>
    <aspect name="com.statful.client.aspects.StatfulAspect"/>
</aspects>

Annotation

After configuring StatfulAspect, you can use @Timer to annotate your methods and measure the time of execution.

@Timer(name = "execution", tags = {"controller", "Entities", "method", "postEntity"})
public Response<Status> postEntity(Entity entity) {
    // ...
}

The above example would send a metric named application.timer.execution with the configured tags and the default timer aggregations.

Quick start

After declaring Statful Client as a dependency, you are ready to use it. The quickest way is to do the following:

// Instantiates a new HTTP client and build the configuration fluently
StatfulClient client = StatfulFactory.buildHTTPClient().with()
        .token("STATFUL_API_TOKEN")
        .build();

// Send a metric
client.counter("transactions", 1).send();

IMPORTANT: This configuration uses the default host and port. You can learn more about configuration in Reference.

Custom Transport

You can add support for your own custom transport, besides the currently supported HTTP and UDP. To adapt the client to your implementation, you must perform two steps:

  1. Implement the TransportSender interface, which defines a method to send metric using your own transport.
  2. Implement a client factory by extending the abstract class CustomStatfulFactory`.

By extending implementing TransportSender buildTransportSender(final ClientConfiguration configuration), your custom TransportSender` implementation has access to all global client configuration, as well as the asynchronous sending of metrics, so you only need to worry about sending metrics.

Custom Transport Example

First implement the TransportSender interface, then you can implement your client factory similar to the following way:

public final class StatfulFactory {

    private static final ThriftClientFactory THRIFT_CLIENT_FACTORY = new ThriftClientFactory();

    private StatfulFactory() { }

    public static StatfulClientBuilder buildThriftClient() {
        // The client is built with default underlying mechanism used in HTTP and UDP transports, including asynchronous sending
        return THRIFT_CLIENT_FACTORY.buildClient();
    }

    private static class ThriftClientFactory extends CustomStatfulFactory {
        ThriftClientFactory() {
            super(Transport.OTHER);
        }

        @Override
        protected TransportSender buildTransportSender(final ClientConfiguration configuration) {
            // Has access to global configuration
            return new ThriftSender(configuration.getHost(), configuration.getPort());
        }
    }
}

To use your code, you need to include the following Statful dependency in your POM.xml file:

<dependency>
    <groupId>com.statful.client</groupId>
    <artifactId>core</artifactId>
    <version>${statful-client.version}</version>
</dependency>

And then, use your own implementation to send metrics:

StatfulClient client = StatfulFactory.buildThriftClient().with()
        .app("AccountService")
        .host("statful-relay.yourcompany.com")
        .tag("cluster", "production")
        .build();

client.gauge("testGauge", 10).send();

HTTP Authentication Token

If you implement HTTP-based transport, you need to send the M-Api-Token header with your token. You can access the token name inside your client factory using ClientConfiguration.TOKEN_HEADER.

Avoiding metrics back-pressure

When the rate of metrics sent to the Statful client is too high, metrics start to be lost and warn log messages are printed. To help reduce this problem there are multiple configuration parameters that can be useful:

  • timeout: Reducing communication timeout, for when the network latency is too high, helps free busy workers sooner to handle other metrics in the buffer.
  • workerPoolSize: Increase the workers pool (default is 1). Note that each worker spawns a thread and, if too many workers are spawn, thread contention could start to be a problem.
  • flushSize and flushInterval: Increasing metrics flush size and interval helps reducing back-pressure but it'll have memory impact and and the payload size sent to Statful will be bigger.

Examples

You can find here some useful usage examples of the Statful Client. In the following examples is assumed you have already included Statful Client in your project.

UDP Configuration

Creates a simple UDP configuration for the client.

StatfulClient client = StatfulFactory.buildUDPClient().with()
        .host("statful-relay.yourcompany.com")
        .app("AccountService")
        .tag("cluster", "production")
        .build();

HTTP Configuration

Creates a simple HTTP API configuration for the client.

StatfulClient client = StatfulFactory.buildHTTPClient().with()
        .app("AccountService")
        .token("STATFUL_API_TOKEN")
        .tag("cluster", "production")
        .build();

Defaults Configuration Per Method

Creates a configuration for the client with custom default options per method.

StatfulClient client = StatfulFactory.buildHTTPClient().with()
        .tag("cluster", "production")
        .token("STATFUL_API_TOKEN")
        .counter(agg(AVG)).counter(aggFreq(FREQ_180))
        .gauge(agg(FIRST)).gauge(aggFreq(FREQ_180))
        .timer(agg(COUNT)).timer(aggFreq(FREQ_180))
        .timer(tag("cluster", "qa"))
        .build();

Mixed Complete Configuration

Creates a configuration defining a value for other available options.

StatfulClient client = StatfulFactory.buildHTTPClient().with()
        .token("STATFUL_API_TOKEN")
        .namespace("application")
        .isDryRun(true)
        .flushInterval(5000)
        .flushSize(50)
        .timeoutMs(300)
        .workerPoolSize(2)
        .connectionTimeoutMs(300)
        .connectionPoolSize(4)
        .sampleRate(10)
        .secure(false)
        .build();

Add metrics

Creates a simple client configuration and use it to send some metrics.

StatfulClient client = StatfulFactory.buildHTTPClient().with()
        .app("AccountService")
        .token("TOKEN")
        .tag("cluster", "production")
        .build();

// Send three different metrics (gauge, timer and a counter)
client.gauge("testGauge", 10).send();
client.timer("testTimer", 100).send();
client.counter("testTimer", 1).send();
client.put("testCustomMetric", 1).send();

// Metric to be sent with tags
client.counter("testCounter", 1).with()
        .tag("host", "localhost").tag("status", "SUCCESS")
        .send();

Add aggregated metrics

Create a simple client configuration and use it to store previously aggregated metrics. This is useful for existing collectors that expose metrics aggregated over periods of time.

StatfulClient client = StatfulFactory.buildHTTPClient().with()
        .app("AccountService")
        .token("TOKEN")
        .tag("cluster", "production")
        .build();

// Send three different metrics (gauge, timer and a counter)
client.aggregatedGauge("testGauge", 10, Aggregation.AVG, AggregationFreq.FREQ_120).send();
client.aggregatedTimer("testTimer", 100, Aggregation.AVG, AggregationFreq.FREQ_10).send();
client.aggregatedCounter("testTimer", 1, Aggregation.SUM, AggregationFreq.FREQ_10).send();
client.aggregatedPut("testCustomMetric", Aggregation.COUNT, AggregationFreq.FREQ_180).send();

// Metric to be sent with tags
client.aggregatedCounter("testCounter", 1, Aggregation.SUM, AggregationFreq.FREQ_30).with()
        .tag("host", "localhost").tag("status", "SUCCESS")
        .send();

Reference

Detailed reference if you want to take full advantage from Statful.

Global configuration

The custom options that can be set on config param are detailed below.

Option Description Type Default Required
app Defines the application global name. If specified sets a global tag app=setValue. String none NO
dryRun Defines if metrics should be output to the logger instead of being sent. Log level is DEBUG. boolean false NO
flushInterval Defines the periodicity of buffer flushes in miliseconds. int 3000 NO
flushSize Defines the maximum buffer size before performing a flush. int 1000 NO
namespace Defines the global namespace. String application NO
sampleRate Defines the rate sampling. Should be a number between [1, 100]. int 100 NO
tags Defines a list global tags. String, String pairs Empty list of tags NO
host Defines the host name to where the metrics should be sent. String api.statful.com NO
port Defines the port. int 443 NO
token Defines the authentication token to be used. String none NO
timeout Defines the timeout for the transport layers in miliseconds. Must be set inside api. long 2000 NO
secure Enable or disables HTTPS. boolean true NO
connectTimeout Connection timeout for http/tcp transports in milliseconds. long 500 NO
connectionPoolSize Connection pool size. int 10 NO
workerPoolSize Asynchronous workers pool size. int 1 NO

Methods

// normal methods
client.counter("testCounter", 1).with().aggregations(SUM).send();
client.gauge("testGauge", 10).with().tag("host", "localhost").send();
client.timer("testTimer", 200).with().namespace("sandbox").send();
client.metric("testCustomMetric").with().namespace("my-namespace").send();
client.counter("transactions").with().tag("host", "localhost").tag("status", "SUCCESS").send();
client.counter("transactions").with().aggregations(AVG, P90).aggFreq(FREQ_10).send();

// aggregated methods
client.aggregatedGauge("testGauge", 10, Aggregation.AVG, AggregationFreq.FREQ_120).send();
client.aggregatedTimer("testTimer", 100, Aggregation.AVG, AggregationFreq.FREQ_10).send();
client.aggregatedCounter("testTimer", 1, Aggregation.SUM, AggregationFreq.FREQ_10).send();
client.aggregatedPut("testCustomMetric", Aggregation.COUNT, AggregationFreq.FREQ_180).send();

The methods for sending metrics receive a metric name and a metric value as arguments and send a counter/gauge/timer/custom metric. Read the methods options reference bellow to get more information about the default values.

Option Description Default for Counter Default for Gauge Default for Timer Default for Custom Metric
aggregations Defines the aggregations to be executed. These aggregations are merged with the ones configured globally, including method defaults.

Valid Aggregations: AVG, COUNT, SUM, FIRST, LAST, P90, P95, P99 MIN, MAX		 SUM, COUNT LAST AVG, P90, COUNT none
aggFreq Defines the aggregation frequency in seconds. It overrides the global aggregation frequency configuration.

Valid Aggregation Frequencies: 10, 30, 60, 120, 180, 300		 10 10 10 10
namespace Defines the namespace of the metric. It overrides the global namespace configuration. application application application application
tags Defines the tags of the metric. These tags are merged with the ones configured globally, including method defaults. none none unit: 'ms' none
timestamp Defines the timestamp of the metric. This timestamp is a POSIX/Epoch time in seconds. current timestamp current timestamp current timestamp current timestamp

Note that calling the aggregations method on the aggregatedX methods will throw UnsupportedOperationException.

Enabling/disabling Statful

// The following metric will be sent to Statful
statful.enable();
statful.counter("transactions").send();

// The following metric will NOT be sent to Statful
statful.disable();
statful.counter("transactions").send();

Authors

Mindera - Software Craft

License

Statful Java Client is available under the MIT license. See the LICENSE file for more information.