As I did as Hadoop engineer, I want to read code and try to change codebase of Presto because it is the shortest way to understand internal architecture and some of tips to use distributed system efficiently.
Presto install instruction can be found official documentation or Teradata documentation. These instructions are written mainly for production usage. However I want to try Presto easily for development usage, which means we can build, launch, test cycle fast.
So that’s the reason why I wrote docker-presto-cluster. docker-presto-cluster was made for same reason of the docker-hadoop-cluster. I want to deploy self-build package onto multi node cluster easily. docker-hadoop-cluster enabled me to do so. I think we can same development style of Presto by using docker-presto-cluster.
The usage is very simple.
$ cd presto-base $ docker build -t lewuathe/presto-base:latest .
$ docker-compose build
$ docker-compose up -d
That’s it. You can see Presto coordinator UI from http://localhost:8080 since docker container expose 8080 port number. One thing to note is that it does not support self-build package deploy. As defined in Dockerfile, it downloads the latest released package. The reason why I did this workaround was that I couldn’t launch Presto cluster with self-build package on OSX due to BSD virtual machine type error. I’m not sure why this error was thrown, but OSX build package does not work on Linux machine like docker container. (Of course you can run OSX build package on OSX)
I’m investigating the reason. At the same time, I’m now preparing docker image to build Presto package which provides us Presto build environment on Linux. After this, we can deploy self-build Presto package more easily.
Thanks.Written on August 6th, 2016 by lewuathe