A Docker image running Elasticsearch on Fedora, tuned for development only.
The official image from Elastic is designed to be flexible and comes with safe defaults, to make sure it's a reasonable starting point for many use cases. For quick integration testing we need mostly small indexes and use "throw away" cheap storage, which leads me to reconfigure most of the official image. Not least I don't need to run the additional plugins it comes with, and no need to waste 2G of heap for the JVM tuning.
To make sure our software behaves as intended, we'll also strictly ban deprecated features and not recommended settings, have the logger spit out a warning for anything dodgy.
Finally, we disable clustering. This image is certainly not intended for complex failure testing nor performance / scalability tests, for any of those use the official images.
The official image is based on CentOS, which is great an reliable to run services, but when developing I prefer to use Fedora so I'm ready for the next generation of CentOS / Red Hat Enterprise Linux. So this is based on Fedora Linux.
docker pull sanne/elasticsearch-light-testing
docker run --ulimit memlock=-1:-1 -it --net=host --rm=true --memory-swappiness=0 --name es-it sanne/elasticsearch-light-testing
Done!
Keep in mind that this image is meant for development in quick iterations; the above options will terminate the container on CTRL+C and remove any data stored.
I highly recommend using tmpfs for storage as Docker is otherwise not very efficient; direct host network usage also helps with efficiency:
docker run --ulimit memlock=-1:-1 -ti --tmpfs /run --tmpfs=/opt/elasticsearch/volatile/data:uid=1000 --tmpfs --tmpfs=/opt/elasticsearch/volatile/logs:uid=1000 --net=host --name es-it sanne/elasticsearch-light-testing
Unfortunately the parameters format Docker requires for tmpfs changes across various versions:
you might need to adapt the above suggestion.
Get rid of it after testing:
docker stop es-it && docker rm es-it
Registered on DockerHub: https://hub.docker.com/r/sanne/elasticsearch-light-testing/
Build it locally:
docker build -t customtest .
Don't forget to change customtest consistently in the run script:
docker run --ulimit memlock=-1:-1 -it --net=host --rm=true --memory-swappiness=0 --name es-it customtest