Provides a Consul based discovery strategy for Hazlecast 3.6+ enabled applications. This is an easy to configure plug-and-play Hazlecast DiscoveryStrategy that will optionally register each of your Hazelcast instances with Consul and enable Hazelcast nodes to dynamically discover one another via Consul.
- Status
- Releases
- Requirements
- Maven/Gradle install
- Features
- Usage
- Build from source
- Unit tests
- Related Info
- Todo
- Notes
- Docker info
- Consul ACL issues?
This is release candidate code, tested against Hazelcast 3.6-EA+ through 3.9.x stable releases, as well as Consul 1.0.x up to 1.2.x.
- MASTER - in progress, this README refers to what is in the master tag. Switch to relevant RELEASE tag above to see that versions README
- Java 6+
- Hazelcast 3.6+
- Consul
To use this discovery strategy in your Maven or Gradle project use the dependency samples below.
repositories {
jcenter()
}
dependencies {
compile 'cc.springcloud:hazelcast-consul-discovery-spi:1.0'
}
<dependencies>
<dependency>
<groupId>cc.springcloud</groupId>
<artifactId>hazelcast-consul-discovery-spi</artifactId>
<version>1.0</version>
</dependency>
</dependencies>
<repositories>
<repository>
<snapshots>
<enabled>false</enabled>
</snapshots>
<id>central</id>
<name>bintray</name>
<url>http://jcenter.bintray.com</url>
</repository>
</repositories>
-
Supports two modes of operation:
- Read-write: peer discovery and registration of a hazelcast instance without a local Consul agent (self registration)
- Read-only: peer discovery only with an existing Consul agent setup (no registration by the strategy itself)
-
If you don't want to use the built in Consul registration, just specify the
DoNothingRegistrator(see below) in your hazelcast discovery-strategy XML config. This will require you to run your own Consul agent that defines the hazelcast service. -
If using self-registration, either
LocalDiscoveryNodeRegistratororExplicitIpPortRegistratorwhich additionally support:- Automatic registration of the hazelcast instance with Consul
- Custom Consul health-check script/interval to validate Hazelcast instance healthly
- Control which IP is published as the service-address with Consul
- Configurable discovery delay
- Automatic Consul de-registration of instance via ShutdownHook
-
Ensure your project has the
hazelcast-consul-discovery-spiartifact dependency declared in your maven pom or gradle build file as described above. Or build the jar yourself and ensure the jar is in your project's classpath. -
Have Consul running and available somewhere on your network, start it such as:
consul agent -server -bootstrap-expect 1 -data-dir /tmp/consul -config-dir /path/to/consul.d/ -ui-dir /path/to/consul-web-ui [-enable-script-checks]
-
Configure your hazelcast.xml configuration file to use the
ConsulDiscoveryStrategy(similar to the below): See hazelcast-consul-discovery-spi-example.xml for a full example with documentation of options. -
Launch your hazelcast instances, configured with the Consul discovery-strategy similar to the below: see ManualRunner.java example.
<network>
<port auto-increment="true">5701</port>
<join>
<multicast enabled="false"/>
<aws enabled="false"/>
<tcp-ip enabled="false" />
<discovery-strategies>
<discovery-strategy enabled="true"
class="cc.springcloud.hazelcast.discovery.consul.ConsulDiscoveryStrategy">
<properties>
<property name="consul-host">localhost</property>
<property name="consul-port">8500</property>
<property name="consul-service-name">hz-discovery-test-cluster</property>
<property name="consul-healthy-only">true</property>
<property name="consul-service-tags">hazelcast, test1</property>
<property name="consul-discovery-delay-ms">10000</property>
<property name="consul-acl-token"></property>
<property name="consul-ssl-enabled">false</property>
<property name="consul-registrator">cc.springcloud.hazelcast.discovery.consul.LocalDiscoveryNodeRegistrator</property>
<property name="consul-registrator-config"><![CDATA[
{
"preferPublicAddress":false,
"Check": {
"Interval": 30s,
"Timeout": 30s,
"TCP": "#MYIP:#MYPORT",
"DeregisterCriticalServiceAfter": null,
"TLSSkipVerify": false,
"Status": null
}
}
]]></property>
</properties>
</discovery-strategy>
</discovery-strategies>
</join>
</network>
- Once nodes are joined you can query Consul to see the auto-registration of hazelcast instances works, the service-id's generated etc
curl http://localhost:8500/v1/catalog/services
{
"consul":[],
"hz-discovery-test-cluster":["hazelcast","test1"],
"web":["rails"]
}
curl http://localhost:8500/v1/catalog/service/hz-discovery-test-cluster
[
{
"Node":"myhost1",
"Address":"192.168.0.208",
"ServiceID":"hz-discovery-test-cluster-192.168.0.208-192.168.0.208-5701",
"ServiceName":"hz-discovery-test-cluster",
"ServiceTags":[
"hazelcast",
"test1"
],
"ServiceAddress":"192.168.0.208",
"ServicePort":5701
},
{
"Node":"myhost1",
"Address":"192.168.0.208",
"ServiceID":"hz-discovery-test-cluster-192.168.0.208-192.168.0.208-5702",
"ServiceName":"hz-discovery-test-cluster",
"ServiceTags":[
"hazelcast",
"test1"
],
"ServiceAddress":"192.168.0.208",
"ServicePort":5702
}
]
-
From the root of this project, build a Jar :
./gradlew assemble -
Include the built jar artifact located at
build/libs/hazelcast-consul-discovery-spi-[VERSION].jarin your hazelcast project -
If not already present in your hazelcast application's Maven (pom.xml) or Gradle (build.gradle) dependencies section; ensure that these dependencies are present (versions may vary as appropriate):
compile group: 'com.ecwid.consul', name: 'consul-api', version:'1.3.1'
compile group: 'com.hazelcast', name:'hazelcast', version:'3.9.4'
Showing LocalDiscoveryNodeRegistrator configured hazelcast services with health-checks
It may also help you to understand the functionality by checking out and running the unit-tests located at src/test/java. BE SURE TO READ the comments in the test source files as some of the tests require you to setup your local Consul and edit certain files.
From the command line you can run TestExplicitIpPortRegistrator and TestLocalDiscoveryNodeRegistrator unit-tests by invoking the runTests task using gradlew that runs both tests and displays the result on the console.
$ ./gradlew runTests
The task above will display output indicating the test has started and whether the test has passed or failed.
cc.springcloud.hazelcast.discovery.consul.TestExplicitIpPortRegistrator > testExplicitIpPortRegistrator STARTED
cc.springcloud.hazelcast.discovery.consul.TestExplicitIpPortRegistrator > testExplicitIpPortRegistrator PASSED
cc.springcloud.hazelcast.discovery.consul.TestDoNothingRegistrator > testDoNothingRegistrator STARTED
cc.springcloud.hazelcast.discovery.consul.TestDoNothingRegistrator > testDoNothingRegistrator FAILED
java.lang.AssertionError at TestDoNothingRegistrator.java:85
To run individual unit-test, use the unitTest.single argument to provide the unit-test you would like to run. The command below runs the unit test for TestDoNothingRegistrator
$ ./gradlew -DunitTest.single=TestDoNothingRegistrator unitTest
The TestDoNothingRegistrator unit-test should be run separately using the unitTest.single argument as demonstrated above as it requires you to register a service with your local consul with 5 nodes/instances. Please CAREFULLY READ the comments in TestDoNothingRegistrator.java to see how this test should be run.
The following parameters can be passed with the -D option when invoking the tests
-DconsulPort=(some port)
-DconsulHost=(some host)
-DconsulAclToken=(some ACL token if the server requires it)
-DconsulSslEnabled=(true | false)
-DconsulSslServerCertFilePath=(/path/to/ca.cert)
-DconsulSslServerCertBase64=(base64 encoded cert string)
-DconsulSslServerHostnameVerify=(false|True)
-DconsulHealthCheckProvider=(cc.springcloud.hazelcast.discovery.consul.ScriptHealthCheckBuilder | cc.springcloud.hazelcast.discovery.consul.HttpHealthCheckBuilder)
- https://www.consul.io
- http://docs.hazelcast.org/docs/3.6/manual/html-single/index.html#discovery-spi
- https://www.consul.io/docs/guides/acl.html#complete-acl-coverage-in-consul-0-8
- Swarm version of this: https://github.com/bitsofinfo/hazelcast-docker-swarm-discovery-spi
- Etcd version of this: https://github.com/bitsofinfo/hazelcast-etcd-discovery-spi
- Ensure all configuration tweakable via
-Dsystem properties
This library may also be helpful to you: docker-discovery-registrator-consul
One of the main drivers for coding this module was for Hazelcast applications that were deployed as Docker containers that would need to automatically register themselves with Consul for higher level cluster orchestration of the cluster.
If you are deploying your Hazelcast application as a Docker container, one helpful tip is that you will want to avoid hardwired
configuration in the hazelcast XML config, but rather have your Docker container take startup arguments that would be translated
to -D system properties on startup. Convienently Hazelcast can consume these JVM system properties and replace variable placeholders in the XML config. See this documentation for examples: http://docs.hazelcast.org/docs/3.6/manual/html-single/index.html#using-variables
Specifically when using this discovery strategy and Docker, it may be useful for you to use the ExplicitIpPortRegistrator ConsulRegistrator instead of the LocalDiscoveryNodeRegistrator as the latter relies on hazelcast to determine its IP/PORT and this may end up being the local container IP, and not the Docker host IP, leading to a situation where a unreachable IP/PORT combination is published to Consul.
Example: excerpt from explicitIpPortRegistrator-example.xml
Start your hazelcast app such as with the below, this would assume that hazelcast is actually reachable via this configuration
via your Docker host and the port mappings that were specified on docker run. (i.e. the IP below would be your docker host/port that is mapped to the actual hazelcast app container and port it exposes for hazelcast).
-
This library may also be helpful to you: docker-discovery-registrator-consul
-
Also see: bitsofinfo/hazelcast-consul-discovery-spi#20 for info on how to do this.
See this Docker issue for related info on detecting mapped ports/ip from within a container
java -jar myHzApp.jar -DregisterWithIpAddress=<dockerHostIp> -DregisterWithPort=<mappedContainerPortOnDockerHost> ....
<property name="consul-registrator-config"><![CDATA[
{
"preferPublicAddress": false,
"Check"：{
"Script":"nc -z #MYIP #MYPORT",
"Interval":30s,
"HTTP":"http://#MYIP:80",
"Timeout":30s,
"TCP":"#MYIP:#MYPORT",
"DeregisterCriticalServiceAfter":null,
"TLSSkipVerify":false,
"Status":null,
"TTL":30s
}
}
]]></property>
Until hazelcast fixes the numerous issues around interfaces/binding etc, you may be better off just running your hz app in a docker swarm and use: https://github.com/bitsofinfo/hazelcast-docker-swarm-discovery-spi for peer to peer hazelcast cluster discovery.
Depending on the health check script you are using: (nc -z #MYIP #MYPORT OR /bin/sh exec 6<>/dev/tcp/#MYIP/#MYPORT || (exit 3) should see something like in your Consul agent monitor when the health-check scripts are running:
> consul monitor --log-level trace
2015/11/20 11:21:39 [DEBUG] agent: check 'service:hz-discovery-test-cluster-192.168.0.208-192.168.0.208-5701' script 'exec 6<>/dev/tcp/192.168.0.208/5701 || (exit 3)' output:
2015/11/20 11:21:39 [DEBUG] agent: Check 'service:hz-discovery-test-cluster-192.168.0.208-192.168.0.208-5701' is passing
Depending on the version of Hazelcast you are using, you may see something like these warnings logged when the health-check script interrogates the hazelcast port and does nothing. You are free to monitor the services any way you wish, or not at all by omitting the healthCheckScript JSON property; see See hazelcast-consul-discovery-spi-example.xml for an example.
Dec 13, 2017 10:25:40 AM com.hazelcast.nio.tcp.TcpIpConnection
INFO: [172.20.10.2]:5702 [hazelcast-consul-discovery-spi] [3.9.1] Connection[id=3, /172.20.10.2:5702->/172.20.10.2:57436, endpoint=null, alive=false, type=NONE] closed. Reason: Connection closed by the other side
Dec 13, 2017 10:25:40 AM com.hazelcast.nio.tcp.TcpIpConnection
INFO: [172.20.10.2]:5702 [hazelcast-consul-discovery-spi] [3.9.1] Connection[id=3, /172.20.10.2:5702->/172.20.10.2:57436, endpoint=null, alive=false, type=NONE] closed. Reason: Connection closed by the other side
OR
Nov 20, 2015 6:57:50 PM com.hazelcast.nio.tcp.SocketAcceptorThread
INFO: [192.168.0.208]:5701 [hazelcast-consul-discovery] [3.6] Accepting socket connection from /192.168.0.208:53495
Nov 20, 2015 6:57:50 PM com.hazelcast.nio.tcp.TcpIpConnectionManager
INFO: [192.168.0.208]:5701 [hazelcast-consul-discovery] [3.6] Established socket connection between /192.168.0.208:5701 and /192.168.0.208:53495
Nov 20, 2015 6:57:50 PM com.hazelcast.nio.tcp.nonblocking.NonBlockingSocketWriter
WARNING: [192.168.0.208]:5701 [hazelcast-consul-discovery] [3.6] SocketWriter is not set, creating SocketWriter with CLUSTER protocol!
Nov 20, 2015 6:57:50 PM com.hazelcast.nio.tcp.TcpIpConnection
INFO: [192.168.0.208]:5701 [hazelcast-consul-discovery] [3.6] Connection [/192.168.0.208:53495] lost. Reason: java.io.EOFException[Could not read protocol type!]
This library was originally developed prior to Consul 0.8, as of 0.8+, changes to the ACL system were made which may require you to grant additional access if your target consul is in default deny mode
See:
- https://www.consul.io/docs/guides/acl.html#complete-acl-coverage-in-consul-0-8
- https://www.consul.io/docs/guides/acl.html#bootstrapping-acls
For examples: For the unit tests to work, create a new client token w/ the following policy, then assign the token to the consul-acl-token setting for this SPI's XML config:
service "" { policy = "write" },
node "" { policy = "write" },
agent "" { policy = "read" }
NOTE! The above sample is just that (a simple sample), in production you may want to re-evaluate and lock down further as needed.