GitHub stream data demo using KairosDB with Cassandra
Latest commit 8b1a447 Nov 14, 2016 @mhausenblas mhausenblas fixes Kairos port

GitHub stream data demo with KairosDB

In this tutorial you'll learn how to set up the time series database KairosDB along with the popular NoSQL database Cassandra on DC/OS. We will use the GitHub API as a stream datasource and build a dashboard using Grafana.

The overall application architecture looks as follows:


Note that above exemplary IP addresses and ports are shown, which varies based on the actual deployment.

This demo assumes you have a DC/OS 1.8 cluster with at least 3 private nodes and 1 public node (the latter for the end-user facing components) up and running as well as the DC/OS CLI installed.

Install Cassandra

First, we need to install Cassandra from the public Universe with the RPC enabled (note the checked start_rpc option):

Cassandra install with RPC enabled

Or, alternatively, you can install Cassandra via the DC/OS CLI:

$ dcos package install --options=copt.json cassandra

With the following content for copt.json:

  "cassandra": {
    "start_rpc": true,
    "rpc_port": 9160,
    "rpc_keepalive": true

Note that it takes a couple of minutes until Cassandra is marked as healthy in the DC/OS dashboard; the Cassandra nodes are available via $DCOS_DASHBOARD/service/cassandra/v1/nodes/connect.

Launch KairosDB

KairosDB is a time series database that runs on top of Cassandra, offering a HTTP data API as well as a Web UI, both exposed via port 8080. Use the DC/OS CLI to launch the KairosDB:

$ dcos marathon app add

Once you see KairosDB running as a service, you can access its Web UI by looking up the IP address of the public agent ( in my case) along with the port that DC/OS has assigned to the container:

$ dcos marathon task list
APP         HEALTHY          STARTED              HOST    ID
/cassandra    True   2016-11-13T05:49:27.871Z  cassandra.ed41b93d-a964-11e6-83aa-9ac7779d790a
/kairos       True   2016-11-13T06:04:13.650Z  kairos.00021191-a967-11e6-83aa-9ac7779d790a
$ dcos marathon task show kairos.00021191-a967-11e6-83aa-9ac7779d790a
  "appId": "/kairos",
  "host": "",
  "id": "kairos.ce522993-c9b6-11e5-bf2e-02181a13a4a7",
  "ipAddresses": [
      "ipAddress": "",
      "protocol": "IPv4"
  "ports": [

The first port (in out example this is 24653) is mapped to container port 8080 and if you combine that with the IP address of the public agent you can now access the KairosDB Web UI:

KairosDB UI

Even now, without any data ingested from GitHub, you can toy around with the internal metrics available. Also, if you're interested in the internals of KairosDB on DC/OS, check out the manual launch notes.

Launch Grafana

We want to build a dashboard with Grafana, plotting the time series data from KairosDB. For that we need to first launch Grafana:

$ dcos marathon app add grafana.json

Note that you can look up Grafana's serving port on the public agent in the same fashion as described above for KairosDB.

Next step is to connect Grafana to KairosDB as a backend, which is supported since v2.1. Start by visiting the Grafana Web UI and authenticating with user:**admin** and password:**admin**. Next, add a new data source as shown below:

Grafana KairosDB

Once you've completed this step you're almost good to go.

Get data from GitHub

In this demo we use GitHub activity as the data source. To ingest data from GitHub into KairosDB, a custom Docker image is used, called the GitHub Fetcher. In a nutshell, it polls$ORG/events with a configurable time interval and uses the KairosDB HTTP API to ingest the datapoints.

We launch the GitHub fetcher with a customized value for KAIROSDB_API value in fetcher.json. The value must point to the IP/port that we found earlier for the KairosDB Web UI. The provided endpoint must not end in a slash:

"env": {
    "KAIROSDB_API": "",
    "GITHUB_ORG": "mesosphere",
    "POLL_INTERVAL": "60"


  • You can customize the Github organization by editing GITHUB_ORG.
  • The period between refreshes can be customized by changing POLL_INTERVAL with a default value of 60 sec.

And now you're ready to launch the GitHub fetcher:

$ dcos marathon app add github-fetcher.json


Once you've gone through the preparation steps and launched both KairosDB and Grafana as well as configured the GitHub fetcher as discussed in the previous section, you're ready to start importing and visualizing Github statistics.

As data is ingested from the GitHub API into Cassandra and available in Grafana you can either create your own dashboards or import the one we've prepared and take it from there:

Grafana dashboard import

The result of the import is the GitHub activity of an organization:

Grafana dashboard

Note that you can play around with the tags to show a drilldown in terms of activities (such as pull requests, comments, etc.) and actors (GitHub users).