Zero Downtime OSGi Micro-services

Guide

Basic OSGi Micro-services

Step 0 - Get the dependencies

Simply execute the bootstrap.sh script to download and install all the necessary dependencies including:

• Apache Karaf 3.0.3
• Apache Kafka 2.10-0.8.2.0

Dependencies are downloaded to an opt/ directory in the project root.

Step 1 - Build the project

mvn clean package

Step 2 - Install the Karaf dependencies

cp zero-downtime-soa-dependencies/target/zero-downtime-soa-dependencies.kar opt/karaf/deploy/

This step pre-installs the necessary Karaf bundles for Cellar, Camel, Hazelcast, and Kafka required by the bundles.

Step 3 - Install the API and Service bundles

cp hello-service-api/target/hello-service-api.jar opt/karaf/deploy/
cp hello-service/target/hello-service.jar opt/karaf/deploy/

Step 4 - Start Karaf

opt/karaf/bin/karaf

Step 5 - Call the API

When the API bundle starts, it finds an available port on the host to bind to the Jetty server to. Check the Karaf log for which port it started on. You should see a message similar to this:

2015-04-04 19:17:22,695 | INFO | raf-3.0.3/deploy | AbstractConnector | 82 - org.eclipse.jetty.aggregate.jetty-all-server - 8.1.15.v20140411 | Started [email protected]:59953

Hint: If you can't find the port, just execute the following command from the Karaf console:

log:tail | grep rest-api

In this case, the Jetty server started on port 59953. Simply make an HTTP GET request to the API endpoint:

curl -XGET localhost:59953/rest-api/sync/John

And you should get back a personalized message.

Distributed OSGi

Step 0 - Create slave Karaf instances from the main Karaf instance

From the Karaf console:

instance:clone root slave-0
instance:start slave-0
instance:clone root slave-1
instance:start slave-1

Cloning the root Karaf instance copies all the current configuration and deployed bundles to the slaves. It's a good way to spin up new instances in real-time to add capacity without needing to re-install everything.

Step 1 - Copy the Cellar config into the root Karaf instance

cp config/karaf/org.apache.karaf.cellar.node.cfg opt/karaf/etc/

The config contained in this project enables the bundle, config, and feature listeners that are responsible for replicating changes from the root Karaf instance to the slaves:

bundle.listener = true
config.listener = true
feature.listener = true

Decoupled & Distributed Routes

Step 0 - Start Zookeeper and Kafka

There are helpful scripts in the root of the project to start both a basic, single node Zookeeper instance and a 2-node Kafka cluster:

./0-start-zookeeper.sh
./1-start-kafka-1.sh
./1-start-kafka-2.sh

Please note, it's important to start Zookeeper first as Kafka is heavily dependent on it.

Step 1 - Call the Kafka-backed API

curl -XGET localhost:59953/rest-api/async/Mary

Documentation

Karaf

Please refer to the Karaf site for current documentation.

Kafka

Please refer to the Kafka site for current documentation.