This repository contains libraries for instrumenting applications in a variety of languages and reporting these metrics to SignalFx. You will need a SignalFx account and organization API token to use those. For more information on SignalFx and to create an account, go to http://www.signalfx.com.
The recommended way to send metrics with java is to use codahale metrics. If you want to send metrics without integrating with codahale, the module signalfx-java will support this.
- Java 6+ with
signalfx-metrics.
<dependency>
<groupId>com.signalfx.public</groupId>
<artifactId>signalfx-codahale</artifactId>
<version>0.0.20</version>
</dependency>$ git clone https://github.com/signalfx/signalfx-java.git
Cloning into 'signalfx-java'...
remote: Counting objects: 930, done.
remote: Compressing objects: 100% (67/67), done.
remote: Total 930 (delta 20), reused 0 (delta 0)
Receiving objects: 100% (930/930), 146.79 KiB | 0 bytes/s, done.
Resolving deltas: 100% (289/289), done.
Checking connectivity... done.
$ cd signalfx-java
$ mvn install
[INFO] Scanning for projects...
...
...
...
[INFO] SignalFx parent .................................. SUCCESS [ 2.483 s]
[INFO] SignalFx Protocol Buffer definitions ............. SUCCESS [ 5.503 s]
[INFO] SignalFx Protobuf Utilities ...................... SUCCESS [ 2.269 s]
[INFO] SignalFx java libraries .......................... SUCCESS [ 3.728 s]
[INFO] Codahale to SignalFx ............................. SUCCESS [ 2.910 s]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 17.120 s
[INFO] ------------------------------------------------------------------------
Codahale metrics are the recommended way of integrating your java code with SignalFx. More information on the Codahale Metrics library can be found on the Codahale Metrics website.
final MetricRegistry metricRegistery = new MetricRegistry();
final SignalFxReporter signalfxReporter = new SignalFxReporter.Builder(
metricRegistery,
"SIGNALFX_AUTH_TOKEN"
).build();
signalfxReporter.start(1, TimeUnit.SECONDS);
final MetricMetadata metricMetadata = signalfxReporter.getMetricMetadata();// This will send the current time in ms to signalfx as a gauge
metricRegistery.register("gauge", new Gauge<Long>() {
public Long getValue() {
return System.currentTimeMillis;
}
});You can add SignalFx specific metadata to codahale metrics by using the MetricMetadata of the reporter. When you use MetricMetadata, rather than register your metric directly with the metricRegistry, you should call the .register() method you get from the call forMetric(). This will construct a unique codahale string for your metric.
// This will send the size of a queue as a gauge, tagging the queue
// with a dimension to describe it
final Queue customerQueue = new ArrayBlockingQueue(100);
metricMetadata.forMetric(new Gauge<Long>() {
@Override
public Long getValue() {
return customerQueue.size();
}
}).withDimension("queue_name", "customer_backlog")
.register(metricRegistery);It is recommended to create your Codahale object as a counter or gauge as a field of your class then use that field to increment values, but if you don't want to maintain this for code cleanliness you can create it on the fly with our builders. For example, if you wanted a timer with the dimension of the store it is from you could use code like this.
Timer t = metricMetadata.forBuilder(MetricBuilder.TIMERS)
.withMetricName("request_time")
.withDimension("storename", "electronics")
.createOrGet(metricRegistery);
Timer.Context c = t.time();
try {
System.out.println("Doing store things");
} finally {
c.close();
}
// Java 7 alternative:
// try (Timer.Context ignored = t.time()) {
// System.out.println("Doing store things");
// }The default source name for metrics is discovered by SourceNameHelper. If you want to override the default behavior, you can pass a third parameter to your Builder and that String is then used as the source. If you are using AWS, we provide a helper to extract your AWS instance ID and use that as the source. For example:
final SignalFxReporter signalfxReporter = new SignalFxReporter.Builder(
metricRegistery,
"SIGNALFX_AUTH_TOKEN",
SourceNameHelper.getAwsInstanceId()
).build();
After setting up a SignalFxReporter, you can use codahale metrics as
you normally would, reported at the frequency configured to the
SignalFxReporter.
You can also interact with our java library directly if you do not want to use Codahale. To do this, you will need to build the metric to send to signalfx manually using protocol buffers.
DataPointReceiverEndpoint dataPointEndpoint = new DataPointEndpoint();
AggregateMetricSender mf =
new AggregateMetricSender("test.SendMetrics",
new HttpDataPointProtobufReceiverFactory(
dataPointEndpoint)
.setVersion(2),
new StaticAuthToken(auth_token),
Collections.<OnSendErrorHandler>singleton(new OnSendErrorHandler() {
@Override
public void handleError(MetricError metricError) {
System.out.println("Unable to POST metrics: " + metricError.getMessage());
}
}));
Collections.<OnSendErrorHandler>emptyList());
try (AggregateMetricSender.Session i = mf.createSession()) {
i.setDatapoint(
SignalFxProtocolBuffers.DataPoint.newBuilder()
.setMetric("curtime")
.setValue(
SignalFxProtocolBuffers.Datum.newBuilder()
.setIntValue(System.currentTimeMillis()))
.addDimensions(
SignalFxProtocolBuffers.Dimension.newBuilder()
.setKey("source")
.setValue("java"))
.build());
}