Go-KV

Simple Distributed in-memory key/value store. GoKV provides high availability and fault tolerance which makes it suitable large-scale applications system without sacrificing performance and reliability. With GoKV, you can instantly create a fast, scalable, distributed system across a cluster of computers.

Installation

go get github.com/tochemey/gokv

Design

Go-KV is designed to distribute key/value pair in a cluster of computers using push-pull anti-entropy method to replicate nodes' state across the cluster. When a data entry is changed on a node the full state of that entry is replicated to other nodes. This approach makes Go-KV eventually consistent. However, at some point in time the cluster will be in complete synchronised state. For frequent state synchronisation one can set the stateSyncInterval value to a low value. The downside of a low value is that it will increase network traffic.

Features

• Built-in client to interact with the cluster via the following apis:
  • Put: create key/value pair that is eventually distributed in the cluster of nodes. The key is a string and the value is a byte array. One can set an expiry to the key.
  • PutProto: to create a key/value pair where the value is a protocol buffer message
  • PutString: to create a key/value pair where the value is a string
  • PutAny: to create a key/value pair with a given Codec to encode the value type.
  • Get: retrieves the value of a given key from the cluster of nodes. This can return a false negative meaning that the key may exist but at the time of checking it is having yet to be replicated in the cluster.
  • GetProto: retrieves a protocol buffer message for a given key. This requires PutProto or Put to be used to set the value.
  • GetString: retrieves a string value for a given key. This requires PutString or Put to be used to set the value.
  • GetAny: retrieves any value type for a given key. This requires PutAny to be used to set the value.
  • List: retrieves the list of key/value pairs in the cluster at a point in time
  • Exists: check the existence of a given key in the cluster. This can return a false negative meaning that the key may exist but at the time of checking it is having yet to be replicated in the cluster.
  • Delete: delete a given key from the cluster. Node only deletes the key they own
• Built-in janitor to remove expired entries. One can set the janitor execution interval. Bearing in mind of the eventual consistency of the Go-KV, one need to set that interval taking into consideration the stateSyncInterval
• Discovery API to implement custom nodes discovery provider. See: discovery
• Comes bundled with some discovery providers that can help you hit the ground running:
  • kubernetes api integration is fully functional
  • nats integration is fully functional
  • static is fully functional and for demo purpose
  • dns is fully functional

Use Cases

• Distributed cache

Example

There is an example on how to use it with NATs here

Builtin Discovery

nats

To use the nats discovery provider one needs to provide the following:

• Server: the NATS Server address
• Subject: the NATS subject to use
• Timeout: the nodes discovery timeout
• MaxJoinAttempts: the maximum number of attempts to connect an existing NATs server. Defaults to 5
• ReconnectWait: the time to backoff after attempting a reconnect to a server that we were already connected to previously. Default to 2 seconds
• DiscoveryPort: the discovery port of the running node
• Host: the host address of the running node

dns

This provider performs nodes discovery based upon the domain name provided. This is very useful when doing local development using docker.

To use the DNS discovery provider one needs to provide the following:

• DomainName: the domain name
• IPv6: it states whether to lookup for IPv6 addresses.

static

This provider performs nodes discovery based upon the list of static hosts addresses. The address of each host is the form of host:port where port is the discovery port.

kubernetes

To get the kubernetes discovery working as expected, the following need to be set in the manifest files:

• Namespace: the kubernetes namespace
• DiscoveryPortName: the discovery port name
• PortName: the client port name. This port is used by the built-in cluster client for the various operations on the key/value pair distributed store
• PodLabels: the POD labels

Make sure to provide the right RBAC settings to be able to access the pods.