100% Open Source, Continuously Train and Deploy Streaming ML and AI Pipelines

Live Demo

• Click here for a live demo.

Note: Do not load any sensitive data into this environment!

Docker Images

• Click here for all Docker images

Related Training and Workshops

• Click here for a related project used for workshops

Setup Kubernetes Cluster

• Follow the instructions here.

Setup Kubernetes Client CLI

• Follow the instructions here.

Clone this Repo including Submodules

git clone --recursive https://github.com/fluxcapacitor/datasticks.com

Pull Latest Tips of Submodules

cd datasticks.com

git submodule update --recursive --remote && git pull --recurse-submodules

Deploy Datasticks to Kubernetes Cluster

./datasticks-up.sh

Get all Service Host/IPs

kubectl get svc -w

(Optional) Setup Friendly CNAMEs in DNS Pointing to Service Host/IPs above

• ie. AWS Route53 REST API, GoDaddy API, etc

Navigate Browser to Apache Host/IP from Above

http://<apache-host-ip>

Advanced Features and Demos

Real-time Topology View of Live Kuberentes Cluster

kubectl describe svc weavescope-app

https://<KUBERNETES-ADMIN-UI-WEAVESCOPE-HOST-IP>

• Note: You can manually scale Spark Workers through WeaveScope

Manually Scale Spark Workers

kubectl scale --replicas=4 rc spark-worker-2-0-1

bash into Live Docker Container

kubectl get pod

kubectl exec <pod-name> -it -- bash -il

• Note: You can manually bash into live Docker containers through WeaveScope

Auto-scale Spark Workers based on CPU Utilization

kubectl autoscale rc spark-worker-2-0-1 --max=4 --cpu-percent=50

Rolling Update of JupyterHub to Increase spark.max.cores and spark.executor.memory

kubectl rolling-update jupyterhub-master -f jupyterhub-rc-2cores-2gb.yaml

Continuous Deploy, Monitor, and Rollback New Spark ML and TensorFlow AI Models

TODO:  Link to jupyter notebook

Continuous, Incremental Training of Spark ML and TensorFlow AI Models from Kafka

TODO:  Link to jupyter noteook

Highly-scalable, Highly-available Model Serving using Battle-tested NetflixOSS Components

TODO:  Link to Hystrix/Turbine dashboard 

Support

• Email help@fluxcapacitor.com for Support!