Instana Java Trace SDK  

Instana automatically instruments well known frameworks for calls coming into a monitored JVM, which we call Entry (also known as Server in OpenTracing), and calls leaving a monitored JVM, which we call Exit (also known as Client). Usually calls pass instrumented code twice: at the start of a request, and at the end. The time elapsed between those two passings is the duration this call took. This duration, along with additional specific metadata is collected in so called Spans.

For the case where Instana does not (yet) know how to instrument a framework, or for monitoring the requests of a very custom application, this SDK can be used.

There is little code needed to enhance monitoring of a custom application. The most common use cases are:

• Adding a new Entry
• Adding a new Exit
• Collecting additional Metadata

Each of those use cases is demonstrated in instana-java-sdk-sample, which contains a Dropwizard based web application, which uses additional custom code. The two classes required are com.instana.sdk.annotation.Span and com.instana.sdk.support.SpanSupport which contain Javadoc explaining their usage.

To enable the SDK in your application, you should add a dependency to the SDK and make sure you package and deploy the SDK jar with your application.

When using maven you can add it with:

<dependency>
  <groupId>com.instana</groupId>
  <artifactId>instana-java-sdk</artifactId>
  <version>1.1.1</version>
</dependency>

To avoid any unwanted interference by scanning for the SDK annotations, the agent requires the following configuration to whitelist packages for SDK usage:

com.instana.plugin.javatrace:
  instrumentation:
    sdk:
      packages:
        - 'com.acme'

By design, the SDK will remain inactive when no Instana agent is monitoring the JVM process. It will also return to inactivity when the Instana agent is stopped. Therefore it is safe to keep it in your application code even when deploying to systems not yet monitored by Instana.

The whole SDK contained in this repository is provided with an MIT License (see LICENSE.md), to allow any use and conflict with strict open source licensing requirements.

Instana Trace Webservice

Using the Trace SDK REST Web Service, it is possible to integrate Instana into any application written in any language. Each running Instana Agent can be used to feed in custom traces, which can be part of automatically captured traces or completely separated. The Agent offers an endpoint which listens on http://localhost:42699/com.instana.plugin.generic.trace and accepts the following JSON via a POST request:

{
  'spanId': <string>,
  'parentId': <string>,
  'traceId': <string>,
  'timestamp': <64 bit long>,
  'duration': <64 bit long>,
  'name' : <string>,
  'type' : <string>,
  'error' : <boolean>,
  'data' : {
    <string> : <string>
  }
}

spanId is an unique identifier for the span. Define the root span of a trace with the same spanId and traceId; define child spans with a unique spanId, the traceId of the root span and the spanId of the span immediately preceding in the hierarchy as parentId. Trace Id, Span Id and Parent Id are 64 bit unique values encoded as hex string like b0789916ff8f319f.

root (spanId=1, traceID=1)
   child (spanId=2,traceId=1, parentId=1)
      child (spanId=3, traceId=1, parentId=2)

timestamp and duration are in milliseconds. timestamp must be the epoch timestamp coordinated to UTC. name can be any string which is used to visualize and group traces and can contain any text, but it is recommended to keep it simple. type is optional, but when given needs to be either ENTRY, EXIT or INTERMEDIATE. data is optional and can contain arbitrary key-value pairs. error is optional and can be set to true to indicate an erroneous span. Behaviour of supplying duplicate keys is unspecified.

The endpoint also accepts a batch of spans, which then need to be given as array:

[
  {
    // span as above
  },
  {
    // span as above
  }
]

Limitations

Adher to the following rate limits for the trace webservice:

• Maximum API calls/sec: 20
• Maximum payload per POST request: A span must not exceed 4 KiB. The request size must not exceed 4 MiB.
• Maximum batch size (spans/array): 1000

FAQ

Is there a limit for the amount of calls from Agent to Backend?

Data transmission between Instana agent and Instana backend depends on a lot of factors. It is done using a persistent HTTP2 connection and highly optimized.

What is the optimal package size (splitted size) to send 50,000 spans with a size about 40 MiB over the Agent to Backend?

Recommended strategy is to buffer spans for up to one second or until 500 spans were collected, then transmit the spans to the agent. An implementation of this transmission strategy can be seen here:

https://github.com/instana/nodejs-sensor/blob/6ec0d469006ad52f4d5fde7218b163e05bf5b2ad/src/tracing/transmission.js

Which requirements are recommended for best Agent performance (cpu, ram, etc.)?

This depends on the host that should be monitored, e.g. the number of Docker containers per host etc.

How should the agent environment be configured?

See https://docs.instana.io/quick_start/agent_configuration/