elasticsearch

README

PLEASE NOTE: This document applies to the HEAD of the source tree only. If you are using a released version of Kubernetes, you almost certainly want the docs that go with that version.

Documentation for specific releases can be found at releases.k8s.io.

Elasticsearch for Kubernetes

This directory contains the source for a Docker image that creates an instance of Elasticsearch 1.5.2 which can be used to automatically form clusters when used with replication controllers. This will not work with the library Elasticsearch image because multicast discovery will not find the other pod IPs needed to form a cluster. This image detects other Elasticsearch pods running in a specified namespace with a given label selector. The detected instances are used to form a list of peer hosts which are used as part of the unicast discovery mechansim for Elasticsearch. The detection of the peer nodes is done by a program which communicates with the Kubernetes API server to get a list of matching Elasticsearch pods. To enable authenticated communication this image needs a secret to be mounted at /etc/apiserver-secret with the basic authentication username and password.

Here is an example replication controller specification that creates 4 instances of Elasticsearch which is in the file music-rc.yaml.

apiVersion: v1
kind: ReplicationController
metadata:
  labels:
    name: music-db
    namespace: mytunes
  name: music-db
spec:
  replicas: 4
  selector:
    name: music-db
  template:
    metadata:
      labels:
         name: music-db
    spec:
      containers:
      - name: es
        image: kubernetes/elasticsearch:1.0
        env:
          - name: "CLUSTER_NAME"
            value: "mytunes-db"
          - name: "SELECTOR"
            value: "name=music-db"
          - name: "NAMESPACE"
            value: "mytunes"
        ports:
        - name: es
          containerPort: 9200
        - name: es-transport
          containerPort: 9300
        volumeMounts:
        - name: apiserver-secret
          mountPath: /etc/apiserver-secret
          readOnly: true
      volumes:
      - name: apiserver-secret
        secret:
          secretName: apiserver-secret

The CLUSTER_NAME variable gives a name to the cluster and allows multiple separate clusters to exist in the same namespace. The SELECTOR variable should be set to a label query that identifies the Elasticsearch nodes that should participate in this cluster. For our example we specify name=music-db to match all pods that have the label name set to the value music-db. The NAMESPACE variable identifies the namespace to be used to search for Elasticsearch pods and this should be the same as the namespace specified for the replication controller (in this case mytunes).

Before creating pods with the replication controller a secret containing the bearer authentication token should be set up. A template is provided in the file apiserver-secret.yaml:

apiVersion: v1
kind: Secret
metadata:
  name: apiserver-secret
  namespace: NAMESPACE
data:
  token: "TOKEN"

Replace NAMESPACE with the actual namespace to be used and TOKEN with the basic64 encoded versions of the bearer token reported by kubectl config view e.g.

$ kubectl config view
...
- name: kubernetes-logging_kubernetes-basic-auth
...
  token: yGlDcMvSZPX4PyP0Q5bHgAYgi1iyEHv2
 ...   
$ echo yGlDcMvSZPX4PyP0Q5bHgAYgi1iyEHv2 | base64
eUdsRGNNdlNaUFg0UHlQMFE1YkhnQVlnaTFpeUVIdjIK=

resulting in the file:

apiVersion: v1
kind: Secret
metadata:
  name: apiserver-secret
  namespace: mytunes
data:
  token: "eUdsRGNNdlNaUFg0UHlQMFE1YkhnQVlnaTFpeUVIdjIK="

which can be used to create the secret in your namespace:

kubectl create -f apiserver-secret.yaml --namespace=mytunes
secrets/apiserver-secret

Now you are ready to create the replication controller which will then create the pods:

$ kubectl create -f music-rc.yaml --namespace=mytunes
replicationcontrollers/music-db

It's also useful to have a service with an load balancer for accessing the Elasticsearch cluster which can be found in the file music-service.yaml.

apiVersion: v1
kind: Service
metadata:
  name: music-server
  namespace: mytunes
  labels:
    name: music-db
spec:
  selector:
    name: music-db
  ports:
  - name: db
    port: 9200
    targetPort: es
  type: LoadBalancer

Let's create the service with an external load balancer:

$ kubectl create -f music-service.yaml --namespace=mytunes
services/music-server

Let's see what we've got:

$ kubectl get pods,rc,services,secrets --namespace=mytunes

NAME             READY     STATUS    RESTARTS   AGE
music-db-cl4hw   1/1       Running   0          27m
music-db-x8dbq   1/1       Running   0          27m
music-db-xkebl   1/1       Running   0          27m
music-db-ycjim   1/1       Running   0          27m
CONTROLLER   CONTAINER(S)   IMAGE(S)                       SELECTOR        REPLICAS
music-db     es             kubernetes/elasticsearch:1.0   name=music-db   4
NAME           LABELS          SELECTOR        IP(S)            PORT(S)
music-server   name=music-db   name=music-db   10.0.45.177      9200/TCP
                                               104.197.12.157
NAME                  TYPE                                      DATA
apiserver-secret      Opaque                                    1

This shows 4 instances of Elasticsearch running. After making sure that port 9200 is accessible for this cluster (e.g. using a firewall rule for Google Compute Engine) we can make queries via the service which will be fielded by the matching Elasticsearch pods.

$ curl 104.197.12.157:9200
{
  "status" : 200,
  "name" : "Warpath",
  "cluster_name" : "mytunes-db",
  "version" : {
    "number" : "1.5.2",
    "build_hash" : "62ff9868b4c8a0c45860bebb259e21980778ab1c",
    "build_timestamp" : "2015-04-27T09:21:06Z",
    "build_snapshot" : false,
    "lucene_version" : "4.10.4"
  },
  "tagline" : "You Know, for Search"
}
$ curl 104.197.12.157:9200
{
  "status" : 200,
  "name" : "Callisto",
  "cluster_name" : "mytunes-db",
  "version" : {
    "number" : "1.5.2",
    "build_hash" : "62ff9868b4c8a0c45860bebb259e21980778ab1c",
    "build_timestamp" : "2015-04-27T09:21:06Z",
    "build_snapshot" : false,
    "lucene_version" : "4.10.4"
  },
  "tagline" : "You Know, for Search"
}

We can query the nodes to confirm that an Elasticsearch cluster has been formed.

$ curl 104.197.12.157:9200/_nodes?pretty=true
{
  "cluster_name" : "mytunes-db",
  "nodes" : {
    "u-KrvywFQmyaH5BulSclsA" : {
      "name" : "Jonas Harrow",
...
        "discovery" : {
          "zen" : {
            "ping" : {
              "unicast" : {
                "hosts" : [ "10.244.2.48", "10.244.0.24", "10.244.3.31", "10.244.1.37" ]
              },
...
      "name" : "Warpath",
...
        "discovery" : {
          "zen" : {
            "ping" : {
              "unicast" : {
                "hosts" : [ "10.244.2.48", "10.244.0.24", "10.244.3.31", "10.244.1.37" ]
              },
...
        "name" : "Callisto",
...
        "discovery" : {
          "zen" : {
            "ping" : {
              "unicast" : {
                "hosts" : [ "10.244.2.48", "10.244.0.24", "10.244.3.31", "10.244.1.37" ]
              },
...
      "name" : "Vapor",
...
        "discovery" : {
          "zen" : {
            "ping" : {
              "unicast" : {
                "hosts" : [ "10.244.2.48", "10.244.0.24", "10.244.3.31", "10.244.1.37" ]
...

Let's ramp up the number of Elasticsearch nodes from 4 to 10:

$ kubectl scale --replicas=10 replicationcontrollers music-db --namespace=mytunes
scaled
$ kubectl get pods --namespace=mytunes
NAME             READY     STATUS    RESTARTS   AGE
music-db-063vy   1/1       Running   0          38s
music-db-5ej4e   1/1       Running   0          38s
music-db-dl43y   1/1       Running   0          38s
music-db-lw1lo   1/1       Running   0          1m
music-db-s8hq2   1/1       Running   0          38s
music-db-t98iw   1/1       Running   0          38s
music-db-u1ru3   1/1       Running   0          38s
music-db-wnss2   1/1       Running   0          1m
music-db-x7j2w   1/1       Running   0          1m
music-db-zjqyv   1/1       Running   0          1m

Let's check to make sure that these 10 nodes are part of the same Elasticsearch cluster:

$ curl 104.197.12.157:9200/_nodes?pretty=true | grep name
"cluster_name" : "mytunes-db",
      "name" : "Killraven",
        "name" : "Killraven",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Tefral the Surveyor",
        "name" : "Tefral the Surveyor",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Jonas Harrow",
        "name" : "Jonas Harrow",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Warpath",
        "name" : "Warpath",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Brute I",
        "name" : "Brute I",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Callisto",
        "name" : "Callisto",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Vapor",
        "name" : "Vapor",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Timeslip",
        "name" : "Timeslip",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Magik",
        "name" : "Magik",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",
      "name" : "Brother Voodoo",
        "name" : "Brother Voodoo",
          "name" : "mytunes-db"
        "vm_name" : "OpenJDK 64-Bit Server VM",
          "name" : "eth0",