0

What's the best ways to run Druid on DC/OS? I see there are 2 possible options; Marathon and Docker. Which one will you recommend? Is there a better way?

Michael Hausenblas
  • 13,162
  • 4
  • 52
  • 66
Abdel Dridi
  • 3
  • 2
  • Are you talking about http://druid.io or something else? – Michael Hausenblas Sep 21 '16 at 11:18
  • 1
    `Docker` is a containerization technology and `Marathon` is a `Mesos Framework` which schedules services (a subset of which is containerized services) across a Mesos cluster. The better comparison would be `Docker Swarm v/s Marathon`, or `Druid in Docker Container v/s Druid in Mesos Containerizer` and so forth on these lines. – psiyumm Sep 21 '16 at 11:52
  • Yes, I am talking about the best way to deploy Druid.io database on Mesos along with other components like spark, kafka... – Abdel Dridi Sep 21 '16 at 14:38

1 Answers1

1

As Druid has quite a number of "moving" parts (5 if I counted correctly, plus ZK and eventually an external Postgres instance), it's not really easy to get it running in a completely automated way.

I started writing a framework (https://github.com/tobilg/druid-framework) by preparing Dockerfiles, but eventually stopped because of the complexity.

If you are ok with starting/managing the components manually, you can probably just run the Docker images via Marathon. But please keep in mind that this is IMO not a production-ready setup...

Tobi
  • 31,405
  • 8
  • 58
  • 90
  • Thanks Tobi! I saw that DCOS has made packages for kafka, Spark, cassandra... I am thinking it will be easier if they create a Druid package that we could deploy using DCOS CLI? – Abdel Dridi Sep 22 '16 at 19:47
  • The DC/OS packages are just frameworks... The work needs to be invested in any case. – Tobi Sep 25 '16 at 13:11