If you are using a released version of Kubernetes, you should refer to the docs that go with that version.
The latest release of this document can be found [here](http://releases.k8s.io/release-1.4/examples/storage/hazelcast/README.md).Documentation for other releases can be found at releases.k8s.io.
The following document describes the development of a cloud native Hazelcast deployment on Kubernetes. When we say cloud native we mean an application which understands that it is running within a cluster manager, and uses this cluster management infrastructure to help implement the application. In particular, in this instance, a custom Hazelcast bootstrapper is used to enable Hazelcast to dynamically discover Hazelcast nodes that have already joined the cluster.
Any topology changes are communicated and handled by Hazelcast nodes themselves.
This document also attempts to describe the core components of Kubernetes: Pods, Services, and Replication Controllers.
This example assumes that you have a Kubernetes cluster installed and running, and that you have installed the kubectl command line tool somewhere in your path. Please see the getting started for installation instructions for your platform.
This is a somewhat long tutorial. If you want to jump straight to the "do it now" commands, please see the tl; dr at the end.
Source is freely available at:
- Hazelcast Discovery - https://github.com/pires/hazelcast-kubernetes-bootstrapper
- Dockerfile - https://github.com/pires/hazelcast-kubernetes
- Docker Trusted Build - https://quay.io/repository/pires/hazelcast-kubernetes
In Kubernetes, the atomic unit of an application is a Pod. A Pod is one or more containers that must be scheduled onto the same host. All containers in a pod share a network namespace, and may optionally share mounted volumes.
In this case, we shall not run a single Hazelcast pod, because the discovery mechanism now relies on a service definition.
In Kubernetes a Service describes a set of Pods that perform the same task. For example, the set of nodes in a Hazelcast cluster. An important use for a Service is to create a load balancer which distributes traffic across members of the set. But a Service can also be used as a standing query which makes a dynamically changing set of Pods available via the Kubernetes API. This is actually how our discovery mechanism works, by relying on the service to discover other Hazelcast pods.
Here is the service description:
apiVersion: v1
kind: Service
metadata:
labels:
name: hazelcast
name: hazelcast
spec:
ports:
- port: 5701
selector:
name: hazelcastThe important thing to note here is the selector. It is a query over labels, that identifies the set of Pods contained by the Service. In this case the selector is name: hazelcast. If you look at the Replication Controller specification below, you'll see that the pod has the corresponding label, so it will be selected for membership in this Service.
Create this service as follows:
$ kubectl create -f examples/storage/hazelcast/hazelcast-service.yamlThe real power of Kubernetes and Hazelcast lies in easily building a replicated, resizable Hazelcast cluster.
In Kubernetes a Replication Controller is responsible for replicating sets of identical pods. Like a Service it has a selector query which identifies the members of it's set. Unlike a Service it also has a desired number of replicas, and it will create or delete Pods to ensure that the number of Pods matches up with it's desired state.
Replication Controllers will "adopt" existing pods that match their selector query, so let's create a Replication Controller with a single replica to adopt our existing Hazelcast Pod.
apiVersion: v1
kind: ReplicationController
metadata:
labels:
name: hazelcast
name: hazelcast
spec:
replicas: 1
selector:
name: hazelcast
template:
metadata:
labels:
name: hazelcast
spec:
containers:
- resources:
limits:
cpu: 0.1
image: quay.io/pires/hazelcast-kubernetes:0.6.1
name: hazelcast
env:
- name: "DNS_DOMAIN"
value: "cluster.local"
- name: POD_NAMESPACE
valueFrom:
fieldRef:
fieldPath: metadata.namespace
ports:
- containerPort: 5701
name: hazelcastThere are a few things to note in this description. First is that we are running the quay.io/pires/hazelcast-kubernetes image, tag 0.5. This is a busybox installation with JRE 8 Update 45. However it also adds a custom application that finds any Hazelcast nodes in the cluster and bootstraps an Hazelcast instance accordingly. The HazelcastDiscoveryController discovers the Kubernetes API Server using the built in Kubernetes discovery service, and then uses the Kubernetes API to find new nodes (more on this later).
You may also note that we tell Kubernetes that the container exposes the hazelcast port. Finally, we tell the cluster manager that we need 1 cpu core.
The bulk of the replication controller config is actually identical to the Hazelcast pod declaration above, it simply gives the controller a recipe to use when creating new pods. The other parts are the selector which contains the controller's selector query, and the replicas parameter which specifies the desired number of replicas, in this case 1.
Last but not least, we set DNS_DOMAIN environment variable according to your Kubernetes clusters DNS configuration.
Create this controller:
$ kubectl create -f examples/storage/hazelcast/hazelcast-controller.yamlAfter the controller provisions successfully the pod, you can query the service endpoints:
$ kubectl get endpoints hazelcast -o json
{
"kind": "Endpoints",
"apiVersion": "v1",
"metadata": {
"name": "hazelcast",
"namespace": "default",
"selfLink": "/api/v1/namespaces/default/endpoints/hazelcast",
"uid": "094e507a-2700-11e5-abbc-080027eae546",
"resourceVersion": "4094",
"creationTimestamp": "2015-07-10T12:34:41Z",
"labels": {
"name": "hazelcast"
}
},
"subsets": [
{
"addresses": [
{
"ip": "10.244.37.3",
"targetRef": {
"kind": "Pod",
"namespace": "default",
"name": "hazelcast-nsyzn",
"uid": "f57eb6b0-2706-11e5-abbc-080027eae546",
"resourceVersion": "4093"
}
}
],
"ports": [
{
"port": 5701,
"protocol": "TCP"
}
]
}
]
}You can see that the Service has found the pod created by the replication controller.
Now it gets even more interesting.
Let's scale our cluster to 2 pods:
$ kubectl scale rc hazelcast --replicas=2Now if you list the pods in your cluster, you should see two hazelcast pods:
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
hazelcast-nanfb 1/1 Running 0 40s
hazelcast-nsyzn 1/1 Running 0 2m
kube-dns-xudrp 3/3 Running 0 1hTo prove that this all works, you can use the log command to examine the logs of one pod, for example:
$ kubectl log hazelcast-nanfb hazelcast
2015-07-10 13:26:34.443 INFO 5 --- [ main] com.github.pires.hazelcast.Application : Starting Application on hazelcast-nanfb with PID 5 (/bootstrapper.jar started by root in /)
2015-07-10 13:26:34.535 INFO 5 --- [ main] s.c.a.AnnotationConfigApplicationContext : Refreshing org.springframework.context.annotation.AnnotationConfigApplicationContext@42cfcf1: startup date [Fri Jul 10 13:26:34 GMT 2015]; root of context hierarchy
2015-07-10 13:26:35.888 INFO 5 --- [ main] o.s.j.e.a.AnnotationMBeanExporter : Registering beans for JMX exposure on startup
2015-07-10 13:26:35.924 INFO 5 --- [ main] c.g.p.h.HazelcastDiscoveryController : Asking k8s registry at https://kubernetes.default.svc.cluster.local..
2015-07-10 13:26:37.259 INFO 5 --- [ main] c.g.p.h.HazelcastDiscoveryController : Found 2 pods running Hazelcast.
2015-07-10 13:26:37.404 INFO 5 --- [ main] c.h.instance.DefaultAddressPicker : [LOCAL] [someGroup] [3.5] Interfaces is disabled, trying to pick one address from TCP-IP config addresses: [10.244.77.3, 10.244.37.3]
2015-07-10 13:26:37.405 INFO 5 --- [ main] c.h.instance.DefaultAddressPicker : [LOCAL] [someGroup] [3.5] Prefer IPv4 stack is true.
2015-07-10 13:26:37.415 INFO 5 --- [ main] c.h.instance.DefaultAddressPicker : [LOCAL] [someGroup] [3.5] Picked Address[10.244.77.3]:5701, using socket ServerSocket[addr=/0:0:0:0:0:0:0:0,localport=5701], bind any local is true
2015-07-10 13:26:37.852 INFO 5 --- [ main] com.hazelcast.spi.OperationService : [10.244.77.3]:5701 [someGroup] [3.5] Backpressure is disabled
2015-07-10 13:26:37.879 INFO 5 --- [ main] c.h.s.i.o.c.ClassicOperationExecutor : [10.244.77.3]:5701 [someGroup] [3.5] Starting with 2 generic operation threads and 2 partition operation threads.
2015-07-10 13:26:38.531 INFO 5 --- [ main] com.hazelcast.system : [10.244.77.3]:5701 [someGroup] [3.5] Hazelcast 3.5 (20150617 - 4270dc6) starting at Address[10.244.77.3]:5701
2015-07-10 13:26:38.532 INFO 5 --- [ main] com.hazelcast.system : [10.244.77.3]:5701 [someGroup] [3.5] Copyright (c) 2008-2015, Hazelcast, Inc. All Rights Reserved.
2015-07-10 13:26:38.533 INFO 5 --- [ main] com.hazelcast.instance.Node : [10.244.77.3]:5701 [someGroup] [3.5] Creating TcpIpJoiner
2015-07-10 13:26:38.534 INFO 5 --- [ main] com.hazelcast.core.LifecycleService : [10.244.77.3]:5701 [someGroup] [3.5] Address[10.244.77.3]:5701 is STARTING
2015-07-10 13:26:38.672 INFO 5 --- [ cached1] com.hazelcast.nio.tcp.SocketConnector : [10.244.77.3]:5701 [someGroup] [3.5] Connecting to /10.244.37.3:5701, timeout: 0, bind-any: true
2015-07-10 13:26:38.683 INFO 5 --- [ cached1] c.h.nio.tcp.TcpIpConnectionManager : [10.244.77.3]:5701 [someGroup] [3.5] Established socket connection between /10.244.77.3:59951
2015-07-10 13:26:45.699 INFO 5 --- [ration.thread-1] com.hazelcast.cluster.ClusterService : [10.244.77.3]:5701 [someGroup] [3.5]
Members [2] {
Member [10.244.37.3]:5701
Member [10.244.77.3]:5701 this
}
2015-07-10 13:26:47.722 INFO 5 --- [ main] com.hazelcast.core.LifecycleService : [10.244.77.3]:5701 [someGroup] [3.5] Address[10.244.77.3]:5701 is STARTED
2015-07-10 13:26:47.723 INFO 5 --- [ main] com.github.pires.hazelcast.Application : Started Application in 13.792 seconds (JVM running for 14.542)Now let's scale our cluster to 4 nodes:
$ kubectl scale rc hazelcast --replicas=4Examine the status again by checking the logs and you should see the 4 members connected.
For those of you who are impatient, here is the summary of the commands we ran in this tutorial.
# create a service to track all hazelcast nodes
kubectl create -f examples/storage/hazelcast/hazelcast-service.yaml
# create a replication controller to replicate hazelcast nodes
kubectl create -f examples/storage/hazelcast/hazelcast-controller.yaml
# scale up to 2 nodes
kubectl scale rc hazelcast --replicas=2
# scale up to 4 nodes
kubectl scale rc hazelcast --replicas=4See here