punxsutawney

An tool that allows for replaying load over and over and over (and over) again.

It uses Mesos/Marathon for running its HTTP clients. It uses Kafka to read traffic data.

Primary components are:

• CLI - CLI to start HttpClients on Mesos (see Main);
• HttpClient - client running on Mesos, that generates HTTP traffic;
• Emitter - emitter (see SampleEmitter), producing traffic to Kafka topic (consumed by HttpClient(s));

Running

In order to run the tool following steps should be executed:

1. Build via ./gradlew jar
2. Start ./punxsutawney.sh
3. Verify app is running via Mesos & Marathon UIs
4. Start emitting traffic to Kafka topic punxsutawney (see SampleEmitter)
5. Monitor stdout (via Mesos UI) of each HttpClient to see rps, failures, etc

Note: if running N HttpClient instances, Kafka topic should contain at least N partitions, in order for HttpClients to be able to consume concurrently.

Traffic generation

HttpClient consumes Kafka messages, containing serialized Requests instances. Each Requests instance is designed to carry a batch of HTTP requests. Emitter should batch 1k - 10k requests together to populate Requests object.