Skip to content

Latest commit

 

History

History
80 lines (52 loc) · 2.62 KB

README.md

File metadata and controls

80 lines (52 loc) · 2.62 KB
Raw

Elasticsearch Changes Feed Plugin

A plugin for elasticsearch which allows a client to create a websocket connection to an elasticsearch node and receive a feed of changes from the database.

Loosely based on https://github.com/derryx/elasticsearch-changes-plugin

Requirements

  • Elasticsearch 1.4 to 2.2
  • Java 7+

Prebuilt Binary Installation

To install run the following command, replacing {version} with the version you want to install, e.g. v2.2.0. Note that the plugin requires additional permissions which will automatically be granted using the below command (remove -b option if you want to manually approve these permissions)

bin/plugin install -b https://github.com/jurgc11/es-change-feed-plugin/releases/download/{version}/es-changes-feed-plugin.zip

Restart elasticsearch.

Build From Source

git clone https://github.com/jurgc11/es-change-feed-plugin.git
cd es-change-feed-plugin
mvn clean install

Limitations

There is a good explanation here of some of the difficulties inherent in a changes feed from elasticsearch: https://groups.google.com/d/msg/elasticsearch/S3fSfr4Cz3g/fyse5X4ofuYJ

This plugin has the following limitations:

  • Only sends live changes - if the client isn't connected at the time the plugin gets the change, it won't be sent
  • Doesn't send the initial state of any documents. You need to use one of the existing APIs for this
  • No handling of misbehaving endpoints beyond what the underlying websocket implementation does see Project Tyrus
  • Client needs to handle failing nodes

Configuration

changes.port

Port the websocket will use. Default is 9400

changes.primaryShardOnly

Whether only changes to the primary shards or changes to all shards are sent. Default is false

changes.listenSource

Which indices, types and documents the plugin will listen to in the format <index_pattern>[/<type_pattern>[/<document_pattern>]].

Each of the above patterns is can be either:

  • * All indices, types or documents
  • id[,id]* List of specific indices, types or documents

So for example:

* will match everything. This is the default value

*/tweet/* will match all tweets in any index

gb,us/user,tweet/* will match all documents of type user or tweet in the gb or us indices

Messages

Below is an example message the client might receive:

{
    "_id": "testdoc",
    "_index": "testidx",
    "_operation": "INDEX",
    "_source": {
        "hello": "world"
    },
    "_timestamp": "2015-10-23T14:22:44.738Z",
    "_type": "testtyp",
    "_version": 1
}