bbuzz17.md

First Time

• Install Docker. If running OSX, Docker 1.13.1 is compatible with the demo. Newer versions might work but some conflicts have been found with Kubernetes.

• Install OpenShift Origin or Minishift. If running OSX, OpenShift Origin 1.4.1 is known to work with Docker version above. The instructions below assume you're using OpenShift Origin.

• Install Maven 3.

• Install Kubetail.

• Install Jupyter via Anaconda.

If doing live demos and running OSX, installing Gas Mask , a hosts file manager, is highly recommended. Gas Mask enables you to easily override host files, making it easy to force routes to be re-routed correctly without requiring external DNS services.

Pre talk

• Check out pre-early17 branch.

git checkout pre-early17

• Adjust hostPath in datagrid/datagrid.yml to point to the correct folder.

• Make sure Docker is running.

• Docker settings: 5 CPU, 8 GB

• Start OpenShift cluster:

oc cluster up --public-hostname=127.0.0.1

• Deploy all components:

./deploy-all.sh

You can follow progress of deployment of Infinispan server pods via:

kubetail -l cluster=datagrid

• Open Chrome and verify all pods are running.

• Start Jupyter, open live-demo.ipynb and verify that the URL returns 0 results:

cd analytics/analytics-jupyter
~/anaconda/bin/jupyter notebook

• Clear Jupyter output by clicking: Cell / All Output / Clear

Real Time Demo

• Start delays.query.continuous.fx.FxApp from IDE and show how no delays are being sent.

• Implement query in delays.query.continuous.ContinuousQueryVerticle:

Query query = qf.from(StationBoard.class)
   .having("entries.delayMin").gt(0L)
   .build();

• Implement publishing delay in delays.query.continuous.ContinuousQueryVerticle:

value.entries.stream()
   .filter(e -> e.delayMin > 0)
   .forEach(e -> {
      publishDelay(key, e);
});

• Add continuous query in delays.query.continuous.ContinuousQueryVerticle:

continuousQuery.addContinuousQueryListener(query, listener);

• Redeploy delays.query.continuous.ContinuousQueryVerticle verticle:

cd real-time/real-time-vertx
mvn clean package -DskipTests=true
oc start-build real-time --from-dir=. --follow

• Start delays.query.continuous.fx.FxApp from IDE and within a few seconds you should see delays appearing.

Analytics Demo

• Go to Jupyter notebook, open live-demo.ipynb and verify that URL returns 0 entries.

• Implement delay ratio task in delays.java.stream.task.DelayRatioTask class:

Map<Integer, Long> totalPerHour = cache.values().stream()
      .collect(
            serialize(() -> Collectors.groupingBy(
                  e -> getHourOfDay(e.departureTs),
                  Collectors.counting()
            )));

Map<Integer, Long> delayedPerHour = cache.values().stream()
      .filter(e -> e.delayMin > 0)
      .collect(
            serialize(() -> Collectors.groupingBy(
                  e -> getHourOfDay(e.departureTs),
                  Collectors.counting()
            )));

• Recompile and redeploy server task:

cd analytics
mvn clean package -pl analytics-server
yes | cp analytics-server/target/analytics-server-1.0-SNAPSHOT.jar ../datagrid/target/analytics-server.jar

• Go to Jupyter notebook and run each cell again of live-demo.ipynb. Value for analytics.size should be 48. The time with biggest ratio of delayed trains should be 2am.