Hazelcast Discovery Plugin for Kubernetes
This repository contains a plugin which provides the automatic Hazelcast member discovery in the Kubernetes environment.
You can use it in your project deployed on Kubernetes in order to make the embedded Hazelcast members discover each other automatically. This plugin is also included in Hazelcast Docker images, Hazelcast Helm Charts, and Hazelcast OpenShift Docker image.
Requirements and Recommendations
- Your Java Runtime Environment must support TLS 1.2 (which is the case for most modern JREs).
- Versions compatibility:
- hazelcast-kubernetes 2.0+ is compatible with hazelcast 4+
- hazelcast-kubernetes 1.3+ is compatible with hazelcast 3.11.x, 3.12.x
- for older hazelcast versions you need to use hazelcast-kubernetes 1.2.x.
- The recommendation is to use StatefulSet for managing Hazelcast PODs; in case of using Deployment (or ReplicationController), the Hazelcast cluster may start with Split Brain (which will anyway re-form to one consistent cluster in a few minutes).
Embedded mode
To use Hazelcast embedded in your application, you need to add the plugin dependency into your Maven/Gradle file.
Then, when you provide hazelcast.yaml
as presented below or an equivalent Java-based configuration, your Hazelcast
instances discover themselves automatically.
Maven
<dependency>
<groupId>com.hazelcast</groupId>
<artifactId>hazelcast-kubernetes</artifactId>
<version>${hazelcast-kubernetes-version}</version>
</dependency>
Gradle
compile group: "com.hazelcast", name: "hazelcast-kubernetes", version: "${hazelcast-kubernetes-version}"
Understanding Discovery Modes
The following table summarizes the differences between the discovery modes: Kubernetes API and DNS Lookup.
Kubernetes API | DNS Lookup | |
---|---|---|
Description | Uses REST calls to Kubernetes Master to fetch IPs of PODs | Uses DNS to resolve IPs of PODs related to the given service |
Pros | Flexible, supports 3 different options: - Hazelcast cluster per service - Hazelcast cluster per multiple services (distinguished by labels) - Hazelcast cluster per namespace |
No additional configuration required, resolving DNS does not require granting any permissions |
Cons | Requires setting up RoleBinding (to allow access to Kubernetes API) | - Limited to headless Cluster IP service - Limited to Hazelcast cluster per service |
Configuration
This plugin supports two different options of how Hazelcast members discover each others:
- Kubernetes API
- DNS Lookup
Kubernetes API
Kubernetes API mode means that each node makes a REST call to Kubernetes Master in order to discover IPs of PODs (with Hazelcast members).
Granting Permissions to use Kubernetes API
Using Kubernetes API requires granting certain permissions. Therefore, you may need to create rbac.yaml
with the following content.
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
name: default-cluster
roleRef:
apiGroup: rbac.authorization.k8s.io
kind: ClusterRole
name: view
subjects:
- kind: ServiceAccount
name: default
namespace: default
Then, apply rbac.yaml
.
$ kubectl apply -f rbac.yaml
Note: You can be even more strict with the permissions and create your own Role. For details, please check the implementation of Hazelcast Helm Chart.
Creating Service
Hazelcast Kubernetes Discovery requires creating a service to PODs where Hazelcast is running. In case of using Kubernetes API mode, the service can be of any type.
kind: Service
metadata:
name: SERVICE-NAME
spec:
type: LoadBalancer
selector:
app: APP-NAME
ports:
- name: hazelcast
port: 5701
Hazelcast Configuration
The second step is to configure the discovery plugin inside hazelcast.yaml
or an equivalent Java-based configuration.
hazelcast:
network:
join:
multicast:
enabled: false
kubernetes:
enabled: true
namespace: MY-KUBERNETES-NAMESPACE
service-name: MY-SERVICE-NAME
config.getNetworkConfig().getJoin().getMulticastConfig().setEnabled(false);
config.getNetworkConfig().getJoin().getKubernetesConfig().setEnabled(true)
.setProperty("namespace", "MY-KUBERNETES-NAMESPACE")
.setProperty("service-name", "MY-SERVICE-NAME");
There are several properties to configure the plugin, all of them are optional.
namespace
: Kubernetes Namespace where Hazelcast is running; if not specified, the value is taken from the environment variablesKUBERNETES_NAMESPACE
orOPENSHIFT_BUILD_NAMESPACE
. If those are not set, the namespace of the POD will be used (retrieved from/var/run/secrets/kubernetes.io/serviceaccount/namespace
).service-name
: service name used to scan only PODs connected to the given service; if not specified, then all PODs in the namespace are checkedservice-label-name
,service-label-value
: service label and value used to tag services that should form the Hazelcast cluster togetherpod-label-name
,pod-label-value
: pod label and value used to tag pods that should form the Hazelcast cluster together.resolve-not-ready-addresses
: if set totrue
, it checks also the addresses of PODs which are not ready;true
by defaultuse-node-name-as-external-address
: if set totrue
, uses the node name to connect to aNodePort
service instead of looking up the external IP using the API;false
by defaultkubernetes-api-retries
: number of retries in case of issues while connecting to Kubernetes API; defaults to3
kubernetes-master
: URL of Kubernetes Master;https://kubernetes.default.svc
by defaultapi-token
: API Token to Kubernetes API; if not specified, the value is taken from the file/var/run/secrets/kubernetes.io/serviceaccount/token
ca-certificate
: CA Certificate for Kubernetes API; if not specified, the value is taken from the file/var/run/secrets/kubernetes.io/serviceaccount/ca.crt
service-port
: endpoint port of the service; if specified with a value greater than0
, it overrides the default;0
by default
You can use one of service-name
,service-label
(service-label-name,service-label-value) and pod-label
(pod-label-name, pod-label-value) based discovery mechanisms, configuring two of them at once does not make sense.
Note: If you don't specify any property at all, then the Hazelcast cluster is formed using all PODs in your current namespace. In other words, you can look at the properties as a grouping feature if you want to have multiple Hazelcast clusters in one namespace.
DNS Lookup
DNS Lookup mode uses a feature of Kubernetes that headless (without cluster IP) services are assigned a DNS record which resolves to the set of IPs of related PODs.
Creating Headless Service
Headless service is a service of type ClusterIP
with the clusterIP
property set to None
.
kind: Service
metadata:
name: SERVICE-NAME
spec:
type: ClusterIP
clusterIP: None
selector:
app: APP-NAME
ports:
- name: hazelcast
port: 5701
Hazelcast Configuration
The Hazelcast configuration to use DNS Lookup looks as follows.
hazelcast:
network:
join:
multicast:
enabled: false
kubernetes:
enabled: true
service-dns: MY-SERVICE-DNS-NAME
config.getNetworkConfig().getJoin().getMulticastConfig().setEnabled(false);
config.getNetworkConfig().getJoin().getKubernetesConfig().setEnabled(true)
.setProperty("service-dns", "MY-SERVICE-DNS-NAME");
There are 2 properties to configure the plugin:
service-dns
(required): service DNS, usually in the form ofSERVICE-NAME.NAMESPACE.svc.cluster.local
service-dns-timeout
(optional): custom time for how long the DNS Lookup is checked
Note: In this README, only YAML configurations are presented, however you can achieve exactly the same effect using XML or Java-based configurations.
Zone Aware
When using ZONE_AWARE
configuration, backups are created in the other availability zone. This feature is available only for the Kubernetes API mode.
Note: Your Kubernetes cluster must orchestrate Hazelcast Member PODs equally between the availability zones, otherwise Zone Aware feature may not work correctly.
YAML Configuration
partition-group:
enabled: true
group-type: ZONE_AWARE
Java-based Configuration
config.getPartitionGroupConfig()
.setEnabled(true)
.setGroupType(MemberGroupType.ZONE_AWARE);
Note the following aspects of ZONE_AWARE
:
- Kubernetes cluster must provide the well-known Kubernetes annotations
- Retrieving Zone Name uses Kubernetes API, so RBAC must be configured as described here
ZONE_AWARE
feature works correctly when Hazelcast members are distributed equally in all zones, so your Kubernetes cluster must orchestrate PODs equally
Note also that retrieving Zone Name assumes that your container's hostname is the same as POD Name, which is almost always true. If you happen to change your hostname in the container, then please define the following environment variable:
env:
- name: POD_NAME
valueFrom:
fieldRef:
fieldPath: metadata.name
Hazelcast Client Configuration
Depending on whether your Hazelcast Client runs inside or outside the Kubernetes cluster, its configuration looks different.
Inside Kubernetes Cluster
If you have a Hazelcast cluster deployed on Kubernetes and you want to configure Hazelcast Client deployed on the same Kubernetes cluster, then the configuration looks very similar to Hazelcast Member.
Here's an example in case of the Kubernetes API mode.
hazelcast-client:
network:
kubernetes:
enabled: true
namespace: MY-KUBERNETES-NAMESPACE
service-name: MY-SERVICE-NAME
clientConfig.getNetworkConfig().getKubernetesConfig().setEnabled(true)
.setProperty("namespace", "MY-KUBERNETES-NAMESPACE")
.setProperty("service-name", "MY-SERVICE-NAME");
Outside Kubernetes Cluster
If your Hazelcast cluster is deployed on Kubernetes, but Hazelcast Client is in a completely different network, then they can connect only through the public Internet network.
To use Hazelcast Dummy Client (<smart-routing>false</smart-routing>
) you don't need any plugin, it's enough to expose your Hazelcast cluster with a LoadBalancer (or NodePort) service and set its IP/port as the TCP/IP Hazelcast Client configuration. Dummy Client, however, compromises the performance, since all the communication happens against a single Hazelcast member.
To configure Hazelcast Smart Client, you need to perform the following steps:
- Expose each Hazelcast Member POD with a separate LoadBalancer or NodePort service (the simplest way to do it is to install Metacontroller and service-per-pod Decorator Controller)
- Configure ServiceAccount with ClusterRole having at least
get
andlist
permissions to the following resources:endpoints
,pods
,nodes
,services
- Use credentials from the created ServiceAccount in the Hazelcast Client configuration (credentials can be fetched with
kubectl get secret <sevice-account-secret> -o jsonpath={.data.token} | base64 --decode
andkubectl get secret <sevice-account-secret> -o jsonpath={.data.ca\\.crt} | base64 --decode
)
hazelcast-client:
network:
kubernetes:
enabled: true
namespace: MY-KUBERNETES-NAMESPACE
service-name: MY-SERVICE-NAME
use-public-ip: true
kubernetes-master: https://35.226.182.228
api-token: eyJhbGciOiJSUzI1NiIsImtpZCI6IiJ9.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.o-j4e-ducrMmQc23xYDnPr6TIyzlAs3pLNAmGLqPe9Vq1mwsxOh3ujcVKR90HAdkfHIF_Sw66qC9hXIDvxfqN_rLXlOKbvTX3gjDrAnyY_93Y3MpmSBj8yR9yHMb4O29a9UIwN5F2_VoCsc0IGumScU_EhPYc9mvEXlwp2bATQOEU-SVAGYPqvVPs9h5wjWZ7WUQa_-RBLMF6KRc9EP2i3c7dPSRVL9ZQ6k6OyUUOVEaPa1tqIxP7vOgx9Tg2C1KmYF5RDrlzrWkhEcjd4BLTiYDKEyaoBff9RqdPYlPwu0YcEH-F7yU8tTDN74KX5jvah3amg_zTiXeNoe5ZFcVdg
ca-certificate: |
-----BEGIN CERTIFICATE-----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-----END CERTIFICATE-----
Note: Hazelcast Client outside Kubernetes cluster works only in the Kubernetes API mode (it does not work in the DNS Lookup mode).
Rolling Upgrade and Scaling
Hazelcast cluster is easily scalable within Kubernetes. You can use the standard kubectl scale
command to change the cluster size. The same applies the rolling upgrade procedure, you can depend on the standard Kubernetes behavior and just update the new version to your Deployment/StatefulSet configurations.
Note however that, by default, Hazelcast does not shutdown gracefully. It means that if you suddenly terminate more than your backup-count
property (1 by default), you may lose the cluster data. To prevent that from happening, set the following properties:
terminationGracePeriodSeconds
: in your StatefulSet (or Deployment) configuration; the value should be high enough to cover the data migration process-Dhazelcast.shutdownhook.policy=GRACEFUL
: in the JVM parameters-Dhazelcast.graceful.shutdown.max.wait
: in the JVM parameters; the value should be high enough to cover the data migration process
Additionally:
- If you use Deployment (not StatefulSet), you need to set your strategy to RollingUpdate and ensure Pods are updated one by one.
- If you upgrade by the minor version, e.g.,
3.11.4 => 3.12
(Enterprise feature), you need to set the following JVM property-Dhazelcast.cluster.version.auto.upgrade.enabled=true
to make sure the cluster version updates automatically.
All these features are already included in Hazelcast Helm Charts.
Plugin Usages
Apart from embedding Hazelcast in your application as described above, there are multiple other scenarios of how to use the Hazelcast Kubernetes plugin.
Docker images
This plugin is included in the official Hazelcast Docker images:
Please check Hazelcast Kubernetes Code Samples for the their usage.
Helm Chart
Hazelcast is available in the form of Helm Chart in a few versions:
- stable/hazelcast - Hazelcast IMDG in the official Helm Chart repository
- hazelcast/hazelcast - Hazelcast IMDG with Management Center
- hazelcast/hazelcast-enterprise - Hazelcast Enterprise with Management Center
- IBM/hazelcast-enterprise - Hazelcast Enterprise (with Management Center) dedicated for IBM ICP and IKS environments
You can also check Hazelcast Helm Charts at Helm Hub.
Kubernetes/OpenShift Operator
Hazelcast is available as Kubernetes/OpenShift Operator:
- hazelcast/hazelcast-operator - official Hazelcast Operator repository
- Operator Hub - official Hazelcast Operator published at Operator Hub
Red Hat OpenShift
The plugin is used to provide the OpenShift integration, please check Hazelcast OpenShift Code Samples for details.
Istio
Hazelcast works correctly with Istio >= 1.1.0 without any need of additional configuration. You can use it in both client-server and embedded modes. Also, the Hazelcast cluster can be either outside or inside the Istio mesh.