Kafka-connect-elasticsearch-source

Kafka Connect Elasticsearch Source: fetch data from elastic-search and sends it to kafka. The connector fetches only new data using a strictly incremental / temporal field (like a timestamp or an incrementing id). It supports dynamic schema and nested objects/ arrays.

Requirements:

• Elasticsearch 6.x and 7.x
• Java >= 8
• Maven

Output data serialization format:

The connector uses kafka-connect schema and structs, that are agnostic regarding the user serialization method (e.g. it might be Avro or json, etc...).

Bugs or new Ideas?

• Issues tracker: https://github.com/DarioBalinzo/kafka-connect-elasticsearch-source/issues
• Feel free to open an issue to discuss new ideas (or propose new solutions with a PR).

Installation:

Compile the project with:

mvn clean package -DskipTests

You can also compile and running both unit and integration tests (docker is mandatory) with:

mvn clean package

Copy the jar with dependencies from the target folder into connect classpath (e.g /usr/share/java/kafka-connect-elasticsearch ) or set plugin.path parameter appropriately.

Example

Using kafka connect in distributed way, a sample config file to fetch my_awesome_index* indices and to produce output topics with es_ prefix:

{       
  "name": "elastic-source",
   "config": {
             "connector.class":"com.github.dariobalinzo.ElasticSourceConnector",
             "tasks.max": "1",
             "es.host" : "localhost",
             "es.port" : "9200",
             "index.prefix" : "my_awesome_index",
             "topic.prefix" : "es_",
             "incrementing.field.name" : "@timestamp"
        }
}

To start the connector with curl:

curl -X POST -H "Content-Type: application/json" --data @config.json http://localhost:8083/connectors | jq

To check the status:

curl localhost:8083/connectors/elastic-source/status | jq

To stop the connector:

curl -X DELETE localhost:8083/connectors/elastic-source | jq

Documentation

Elasticsearch Configuration

es.host ElasticSearch host. Optionally it is possible to specify many hosts using ; as separator (host1;host2;host3)

• Type: string
• Importance: high
• Dependents: index.prefix

es.port ElasticSearch port

• Type: string
• Importance: high
• Dependents: index.prefix

es.scheme ElasticSearch scheme (http/https)

• Type: string
• Importance: medium
• Default: http

es.user Elasticsearch username

• Type: string
• Default: null
• Importance: high

es.password Elasticsearch password

• Type: password
• Default: null
• Importance: high

connection.attempts Maximum number of attempts to retrieve a valid Elasticsearch connection.

• Type: int
• Default: 3
• Importance: low

connection.backoff.ms Backoff time in milliseconds between connection attempts.

• Type: long
• Default: 10000
• Importance: low

index.prefix Indices prefix to include in copying.

• Type: string
• Default: ""
• Importance: medium

Connector Configuration

poll.interval.ms Frequency in ms to poll for new data in each index.

• Type: int
• Default: 5000
• Importance: high

batch.max.rows Maximum number of documents to include in a single batch when polling for new data.

• Type: int
• Default: 10000
• Importance: low

topic.prefix Prefix to prepend to index names to generate the name of the Kafka topic to publish data

• Type: string
• Importance: high

filters.whitelist Whitelist filter for extracting a subset of fields from elastic-search json documents. The whitelist filter supports nested fields. To provide multiple fields use ; as separator (e.g. customer;order.qty;order.price).

• Type: string
• Importance: medium
• Default: null

filters.json_cast This filter casts nested fields to json string, avoiding parsing recursively as kafka connect-schema. The json-cast filter supports nested fields. To provide multiple fields use ; as separator (e.g. customer;order.qty;order.price).

• Type: string
• Importance: medium
• Default: null

fieldname_converter Configuring which field name converter should be used (allowed values: avro or nop). By default, the avro field name converter renames the json fields non respecting the avro specifications (https://avro.apache.org/docs/current/spec.html#names) in order to be serialized correctly. To disable the field name conversion set this parameter to nop.

• Type: string
• Importance: medium
• Default: avro