ElasticsearchElixirBulkProcessor

Elasticsearch Elixir Bulk Processor is a configurable manager for efficiently inserting data into Elasticsearch. This processor uses GenStages for handling backpressure, and various settings to control the bulk payloads being uploaded to Elasticsearch.

Inspired by the Java Bulk Processor. Uses elasticsearch-elixir as the client. Featured on the Elastic Community Contributed Clients page.

Installation

If available in Hex, the package can be installed by adding elasticsearch_elixir_bulk_processor to your list of dependencies in mix.exs:

def deps do
  [
    {:elasticsearch_elixir_bulk_processor, "~> 0.1"}
  ]
end

Sending data

ElasticsearchElixirBulkProcessor.send_requests(list_of_items)

To send a list of request items to Elasticsearch. This mechanism uses GenStages for back pressure. NOTE: It should be completely reasonable to use this function by passing single element lists, the mechanism aggregates the items together prior to sending them.

If you wish to bypass the GenStage mechanism and send the data synchronously you can use:

ElasticsearchElixirBulkProcessor.Bulk.DirectUpload.add_requests(list_of_items)

The list elements must be structs:

• ElasticsearchElixirBulkProcessor.Items.Index
• ElasticsearchElixirBulkProcessor.Items.Create
• ElasticsearchElixirBulkProcessor.Items.Update
• ElasticsearchElixirBulkProcessor.Items.Delete

Examples

    iex> alias ElasticsearchElixirBulkProcessor.Items.Index
    ...> [
    ...>  %Index{index: "test_index", source: %{"field" => "value1"}},
    ...>  %Index{index: "test_index", source: %{"field" => "value2"}},
    ...>  %Index{index: "test_index", source: %{"field" => "value3"}}
    ...> ]
    ...> |> ElasticsearchElixirBulkProcessor.send_requests()
    :ok

Configuration

Elasticsearch endpoint

Can be configurate via the ELASTICSEARCH_URL environment variable, defaults to: "http://localhost:9200".

config :elasticsearch_elixir_bulk_processor,
       ElasticsearchElixirBulkProcessor.ElasticsearchCluster,
       url: {:system, "ELASTICSEARCH_URL"},
       api: Elasticsearch.API.HTTP

Alternatively:

config :elasticsearch_elixir_bulk_processor,
       ElasticsearchElixirBulkProcessor.ElasticsearchCluster,
       url: "http://localhost:9200",
       api: Elasticsearch.API.HTTP

See the client configuration for more.

Action count

Number of actions/items to send per bulk (can be changed at run time), deault is nil (unlimited):

ElasticsearchElixirBulkProcessor.set_event_count_threshold(100)

Byte size

Max number of bytes to send per bulk (can be changed at run time), default is 62_914_560 (60mb):

ElasticsearchElixirBulkProcessor.set_byte_threshold(100)

Action order

Preservation of order of actions/items

config :elasticsearch_elixir_bulk_processor, preserve_event_order: false

Retries

Retry policy, this uses the ElixirRetry DSL. See ElasticsearchElixirBulkProcessor.Bulk.Retry.policy.

config :elasticsearch_elixir_bulk_processor, retry_function: &MyApp.Retry.policy/0

Default:

def default do
  constant_backoff(100) |> Stream.take(5)
end

Success and error handlers

The callbacks on a successful upload or in case of failed items or failed request can bet set through the config. On success, the handler is called with the Elasticsearch bulk request. On failure, the hanlder is called with%{data: any, error: any}, data being the original payload and error being the response or HTTP error. See ElasticsearchElixirBulkProcessor.Bulk.Handlers.

config :elasticsearch_elixir_bulk_processor,
  success_function: &MyApp.success_handler/1,
  error_function: &MyApp.error_handler/1

It's highly recommended you add an error handler to make sure your data is uploaded succesfully, for example you can use the logger:

  require Logger
  ...

  def error_handler(%{data: _, error: {_, error}}) do
    error
    |> inspect
    |> Logger.error()
  end

Documentation can be generated with ExDoc and published on HexDocs. Once published, the docs can be found at https://hexdocs.pm/elasticsearch_elixir_bulk_processor.

Testing script

The testing script is used to compare insertion using direct upload vs using a GenStage based approach. Run Elasticsearch set up with:

docker-compose up

Run test:

mix insert_test <INSERT_COUNT> <BULK_SIZE> staged|direct

• staged - uses a GenStage mechanism to aggregate and insert.
• direct - iterates and inserts bulk sequentially as given.