waterdudu / docker-elasticsearch

Dockerfile for a base Elasticsearch image to be extended by others (allow to install plug-ins, change configuration, etc.)

docker-elasticsearch

Ready to use, lean and highly configurable Elasticsearch container image.

Current software

• OpenJDK JRE 8u151
• Elasticsearch 6.2.2

Run

Attention

• In order for bootstrap.mlockall to work, ulimit must be allowed to run in the container. Run with --privileged to enable this.
• Multicast discovery is no longer built-in

Ready to use node for cluster elasticsearch-default:

docker run --name elasticsearch \
	--detach \
	--privileged \
	--volume /path/to/data_folder:/data \
        quay.io/pires/docker-elasticsearch:6.2.2

Ready to use node for cluster myclustername:

docker run --name elasticsearch \
	--detach \
	--privileged \
	--volume /path/to/data_folder:/data \
	-e CLUSTER_NAME=myclustername \
        quay.io/pires/docker-elasticsearch:6.2.2

Ready to use node for cluster elasticsearch-default, with 8GB heap allocated to Elasticsearch:

docker run --name elasticsearch \
	--detach \
	--privileged \
	--volume /path/to/data_folder:/data \
	-e ES_JAVA_OPTS="-Xms8g -Xmx8g" \
        quay.io/pires/docker-elasticsearch:6.2.2

Ready to use node with plugins (x-pack and repository-gcs) pre installed. Already installed plugins are ignored:

docker run --name elasticsearch \
	--detach \
	--privileged \
	--volume /path/to/data_folder:/data \
	-e ES_JAVA_OPTS="-Xms8g -Xmx8g" \
	-e ES_PLUGINS_INSTALL="repository-gcs,x-pack" \
        quay.io/pires/docker-elasticsearch:6.2.2

Master-only node for cluster elasticsearch-default:

docker run --name elasticsearch \
	--detach \
	--privileged \
	--volume /path/to/data_folder:/data \
	-e NODE_DATA=false \
	-e HTTP_ENABLE=false \
        quay.io/pires/docker-elasticsearch:6.2.2

Data-only node for cluster elasticsearch-default:

docker run --name elasticsearch \
	--detach --volume /path/to/data_folder:/data \
	--privileged \
	-e NODE_MASTER=false \
	-e HTTP_ENABLE=false \
        quay.io/pires/docker-elasticsearch:6.2.2

Data-only node for cluster elasticsearch-default with shard allocation awareness:

docker run --name elasticsearch \
	--detach --volume /path/to/data_folder:/data \
        --volume /etc/hostname:/dockerhost \
	--privileged \
	-e NODE_MASTER=false \
	-e HTTP_ENABLE=false \
    -e SHARD_ALLOCATION_AWARENESS=dockerhostname \
    -e SHARD_ALLOCATION_AWARENESS_ATTR="/dockerhost" \
        quay.io/pires/docker-elasticsearch:6.2.2

Client-only node for cluster elasticsearch-default:

docker run --name elasticsearch \
	--detach \
	--privileged \
	--volume /path/to/data_folder:/data \
	-e NODE_MASTER=false \
	-e NODE_DATA=false \
        quay.io/pires/docker-elasticsearch:6.2.2

I also make available special images and instructions for AWS EC2 and Kubernetes.

Environment variables

This image can be configured by means of environment variables, that one can set on a Deployment.

• CLUSTER_NAME
• NODE_NAME
• NODE_MASTER
• NODE_DATA
• NETWORK_HOST
• HTTP_ENABLE
• HTTP_CORS_ENABLE
• HTTP_CORS_ALLOW_ORIGIN
• NUMBER_OF_MASTERS
• MAX_LOCAL_STORAGE_NODES
• ES_JAVA_OPTS
• ES_PLUGINS_INSTALL - comma separated list of Elasticsearch plugins to be installed. Example: ES_PLUGINS_INSTALL="repository-gcs,x-pack"
• SHARD_ALLOCATION_AWARENESS
• SHARD_ALLOCATION_AWARENESS_ATTR
• MEMORY_LOCK - memory locking control - enable to prevent swap (default = true) .
• REPO_LOCATIONS - list of registered repository locations. For example ["/backup"] (default = []).
• PROCESSORS - allow elasticsearch to optimize for the actual number of available cpus (must be an integer - default = 1)

Backup

Mount a shared folder (for example via NFS) to /backup and make sure the elasticsearch user has write access. Then, set the REPO_LOCATIONS environment variable to ["/backup"] and create a backup repository:

backup_repository.json:

{
  "type": "fs",
  "settings": {
    "location": "/backup",
    "compress": true
  }
}

curl -XPOST http://<container_ip>:9200/_snapshot/nas_repository -d @backup_repository.json`

Now, you can take snapshots using:

curl -f -XPUT "http://<container_ip>:9200/_snapshot/nas_repository/snapshot_`date --utc +%Y_%m_%dt%H_%M`?wait_for_completion=true"