hekate's Introduction

Hekate.io

Java Library for Distributed Services

License

Open source Apache License v2.0

Features

Distributed Service Discovery
- Decentralized Service Discovery (based on performance-optimized Gossip protocol)
- Easy integration into existing infrastructure
  - Clouds (based on Apache JClouds)
    - Amazon EC2 and S3
    - Google Cloud Storage and Compute Engine
    - Azure Blob Storage
    - etc
  - Kubernetes
  - ZooKeeper
  - Etcd
  - Consul
  - IP Multicast
  - Shared Database (JDBC-based)
  - Shared File System
- User-defined Service Properties and Roles
- Cluster Event Listeners
- Service Topology Views and Filtering APIs
- Health Monitoring and Split-brain Detection
Messaging
- Synchronous and Asynchronous Messaging (backed by Netty)
- Cluster-aware Load Balancing and Routing
- SSL/TLS Encryption of Network Communications (optional)
- Back Pressure Policies
- Error Retry Policies
- Pluggable Serialization
  - Google Protocol Buffers (work in progress)
  - JDK Serialization
  - Manual Serialization
Remote Procedure Calls (RPC)
- Type-safe Invocation of Remote Java Services
- Automatic Discovery and Load Balancing
- Synchronous and Asynchronous APIs
- Multi-node Broadcasting and Aggregation of Results
- Back Pressure Policies
- Error Retry Policies
- ...and everything from the "Messaging" section above:)
Cluster-wide Singleton Service (aka Leader Election )
- Decentralized Leader Election
- Followers are Aware of the Current Leader
- Leader can Dynamically Yield Leadership
Distributed Locks
- Synchronous and Asynchronous Reentrant Locks
- Decentralized Lock Management
- Configurable Lock Groups (aka Lock Regions)
Spring Boot/Framework Support (optional)
- Spring Boot Auto-configurations and @Annotations
- Spring-compliant Beans
- Spring XML Schema
Metrics
- Internal Metrics Recording and Publishing via Micrometer.io
Raft-based Replicated State Machines (work in progress)

Documentation

Javadocs (Follow ~~the White Rabbit~~ [« start here] marks)
Reference Guide (coming soon...)
Spring Boot Properties Reference

Code Examples

Quickstart for Standalone Java Application

public class MyApplication{
    public static void main(String[] args) throws Exception {
        Hekate hekate = new HekateBootstrap()
            .withNamespace("my-cluster")
            .withNodeName("my-node")
            .join();
        
        System.out.println("Cluster topology: " + hekate.cluster().topology());
    }
}

Quickstart for Spring Boot Application

@EnableHekate // <-- Enable Hekate integration.
@SpringBootApplication
public class MyApplication {
    @Autowired
    private Hekate hekate;

    public static void main(String[] args) {
        SpringApplication.run(MyApplication.class, args);
    }
    
    @PostConstruct
    public void start(){
        System.out.println("Cluster topology: " + hekate.cluster().topology());
    }
}

More Examples:

Please see the hekate-io/hekate-examples project for more examples.

Maven artifacts

For projects based on Spring Boot:

<dependency>
    <groupId>io.hekate</groupId>
    <artifactId>hekate-spring-boot</artifactId>
    <version>4.1.3</version>
</dependency>

For projects based on Spring Framework:

<dependency>
    <groupId>io.hekate</groupId>
    <artifactId>hekate-spring</artifactId>
    <version>4.1.3</version>
</dependency>

For standalone applications:

<dependency>
    <groupId>io.hekate</groupId>
    <artifactId>hekate-core</artifactId>
    <version>4.1.3</version>
</dependency>

Other artifacts:
- Cluster Bootstrapping (seed node discovery)
  - hekate-jclouds-core - Integration with Apache JClouds for cloud environments.
  - hekate-jclouds-aws - Extended integration with Amazon EC2 cloud.
  - hekate-kubernetes - Integration with Kubernetes
  - hekate-zookeeper - Integration with Apache ZooKeeper
  - hekate-etcd - Integration with Etcd
  - hekate-consul - Integration with Consul

How to build

Software requirements:

Latest stable Java SDK (8, 11, 17)
Latest stable Docker (required for tests only)

Build (no tests):

cd to the project's root folder
run ./mvnw clean package -DskipTests=true

Build (with tests):

cd to the project's root folder
make a copy of test.properties file with name my_test.properties
edit my_test.properties according to your environment
run docker-compose up -d to prepare Docker-based testing infrastructure
run ./mvnw clean package

Release History

v.4.1.3 (25-Nov-2022) - [release notes]
v.4.1.2 (12-May-2022) - [release notes]
v.4.1.1 (11-May-2022) - [release notes]
v.4.1.0 (3-Apr-2022) - [release notes]
v.4.0.0 (20-Mar-2021) - [release notes]
v.3.11.0 (24-Oct-2020) - [release notes]
v.3.10.0 (4-Jul-2020) - [release notes]
v.3.9.0 (8-May-2020) - [release notes]
v.3.8.0 (29-Feb-2020) - [release notes]
v.3.7.0 (3-Feb-2020) - [release notes]
v.3.6.0 (13-Oct-2019) - [release notes]
v.3.5.0 (14-Aug-2019) - [release notes]
v.3.4.1 (4-Jul-2019) - [release notes]
v.3.4.0 (29-Jun-2019) - [release notes]
v.3.3.0 (19-Apr-2019) - [release notes]
v.3.2.0 (5-Apr-2019) - [release notes]
v.3.1.0 (28-Mar-2019) - [release notes]
v.3.0.0 (19-Mar-2019) - [release notes]
v.2.6.0 (22-Aug-2018) - [release notes]
v.2.5.0 (4-Aug-2018) - [release notes]
v.2.4.1 (15-Jul-2018) - [release notes]
v.2.4.0 (15-Jul-2018) - [release notes]
...and so on

hekate's People

Contributors

Stargazers

Watchers

hekate's Issues

Node fails to start due to unreachable seed node

I love this library, and have started to use it for our clustering need. Lately I noticed an issue that I couldn't find a workaround.

We are using JDBC as seed node provider. I understand each node adds itself to the database table cluster_node when it joins a cluster and removes itself from the table when it leaves the cluster. In addition, there is a cleanup-interval used by the coordinator node to remove inactive nodes.

Occasionally when all the nodes abruptly die together and fail to clean up the table, the table is left with dead nodes. If I start a new node, it tries to reach these dead "seed nodes" and falls into an infinite retry loop.

The exact sequence of events is non-trivial to reproduce, but the same situation can be easily simulated like the following.

Before start any node, manually add a fake node to the table to simulate a dead node.

This fake node is unreachable, now start a real node 192.168.86.129, and watch that it falls in an infinite loop trying to reach the unreachable node.

INFO  2022-05-09 23:38:56,956 main HekateNode 
      Hekate 4.0.0 (build: 20.03.2021, revision: 87b0f48e2d962d000b5c3108fec35e54e6c8b558).
INFO  2022-05-09 23:38:56,960 HekateSys-0-0 NettyNetworkService 
      Network address selected [address=/192.168.86.129, selector=any-ip4]
INFO  2022-05-09 23:38:56,961 HekateSys-0-0 NettyNetworkService 
      Binding network acceptor [port=10012]
INFO  2022-05-09 23:38:57,057 HekateNioAcceptor-2-0 NettyNetworkService 
      Couldn't bind on port 10012 ...will try next port [next-port=10013]
INFO  2022-05-09 23:38:57,061 HekateNioAcceptor-2-0 NettyNetworkService 
      Done binding [bind-address=/0:0:0:0:0:0:0:0:10013]
INFO  2022-05-09 23:38:57,083 HekateSys-0-0 NettyNetworkService 
      Started accepting network connections [address=/0:0:0:0:0:0:0:0:10013]
INFO  2022-05-09 23:38:57,085 HekateCluster-7-0 DefaultClusterService 
      Joining cluster [namespace=model-builder, local-node=/192.168.86.129:10013:f92ee540e26b405090441ebd39d81e75]
INFO  2022-05-09 23:38:57,086 HekateCluster-7-0 HikariDataSource 
      hekateDataSource - Starting...
INFO  2022-05-09 23:38:57,099 HekateCluster-7-0 HikariDataSource 
      hekateDataSource - Start completed.
INFO  2022-05-09 23:38:57,112 HekateClusterGossip-6-0 GossipSeedNodesSate 
      Discovered new seed node [address=SeedNodeState[address=/192.168.0.1:8080, status=NEW]]
INFO  2022-05-09 23:38:57,114 HekateClusterGossip-6-0 GossipSeedNodesSate 
      Discovered new seed node [address=SeedNodeState[address=/192.168.86.129:10012, status=NEW]]
INFO  2022-05-09 23:38:57,115 HekateClusterGossip-6-0 DefaultClusterService 
      Sending cluster join request [seed-node=/192.168.0.1:8080].
INFO  2022-05-09 23:38:58,118 HekateClusterGossip-6-0 DefaultClusterService 
      Sending cluster join request [seed-node=/192.168.86.129:10012].
INFO  2022-05-09 23:38:59,123 HekateClusterGossip-6-0 DefaultClusterService 
      Sending cluster join request [seed-node=/192.168.86.129:10012].
INFO  2022-05-09 23:39:00,130 HekateClusterGossip-6-0 DefaultClusterService 
      Sending cluster join request [seed-node=/192.168.86.129:10012].
WARN  2022-05-09 23:39:00,150 HekateClusterGossip-6-0 GossipSeedNodesSate 
      Seed node timeout ...will retry [address=/192.168.0.1:8080, cause=io.hekate.network.NetworkConnectTimeoutException: connection timed out: /192.168.0.1:8080]
WARN  2022-05-09 23:39:01,124 HekateClusterGossip-6-0 GossipSeedNodesSate 
      Seed node timeout ...will retry [address=/192.168.0.1:8080, cause=io.hekate.network.NetworkConnectTimeoutException: connection timed out: /192.168.0.1:8080]
INFO  2022-05-09 23:39:01,134 HekateClusterGossip-6-0 DefaultClusterService 
      Sending cluster join request [seed-node=/192.168.86.129:10012].
WARN  2022-05-09 23:39:02,128 HekateClusterGossip-6-0 GossipSeedNodesSate 
      Seed node timeout ...will retry [address=/192.168.0.1:8080, cause=io.hekate.network.NetworkConnectTimeoutException: connection timed out: /192.168.0.1:8080]
INFO  2022-05-09 23:39:02,169 HekateClusterGossip-6-0 DefaultClusterService 
      Sending cluster join request [seed-node=/192.168.0.1:8080].
WARN  2022-05-09 23:39:03,132 HekateClusterGossip-6-0 GossipSeedNodesSate 
      Seed node timeout ...will retry [address=/192.168.0.1:8080, cause=io.hekate.network.NetworkConnectTimeoutException: connection timed out: /192.168.0.1:8080]
WARN  2022-05-09 23:39:03,154 HekateClusterGossip-6-0 GossipSeedNodesSate 
      Seed node timeout ...will retry [address=/192.168.0.1:8080, cause=io.hekate.network.NetworkConnectTimeoutException: connection timed out: /192.168.0.1:8080]

Recommend Projects

hekate-io / hekate Goto Github PK