Elasticsearch Connector

This connector provides sinks that can request document actions to an Elasticsearch Index. To use this connector, add one of the following dependencies to your project, depending on the version of the Elasticsearch installation:

Maven DependencySupported sinceElasticsearch version
flink-connector-elasticsearch5_2.111.3.05.x
flink-connector-elasticsearch6_2.111.6.06.x
flink-connector-elasticsearch7_2.111.10.07 and later versions

Note that the streaming connectors are currently not part of the binary distribution. See here for information about how to package the program with the libraries for cluster execution.

Installing Elasticsearch

Instructions for setting up an Elasticsearch cluster can be found here. Make sure to set and remember a cluster name. This must be set when creating an ElasticsearchSink for requesting document actions against your cluster.

Elasticsearch Sink

The ElasticsearchSink uses a TransportClient (before 6.x) or RestHighLevelClient (starting with 6.x) to communicate with an Elasticsearch cluster.

The example below shows how to configure and create a sink:

  1. import org.apache.flink.api.common.functions.RuntimeContext;
  2. import org.apache.flink.streaming.api.datastream.DataStream;
  3. import org.apache.flink.streaming.connectors.elasticsearch.ElasticsearchSinkFunction;
  4. import org.apache.flink.streaming.connectors.elasticsearch.RequestIndexer;
  5. import org.apache.flink.streaming.connectors.elasticsearch5.ElasticsearchSink;
  7. import org.elasticsearch.action.index.IndexRequest;
  8. import org.elasticsearch.client.Requests;
  10. import java.net.InetAddress;
  11. import java.net.InetSocketAddress;
  12. import java.util.ArrayList;
  13. import java.util.HashMap;
  14. import java.util.List;
  15. import java.util.Map;
  17. DataStream<String> input = ...;
  19. Map<String, String> config = new HashMap<>();
  20. config.put("cluster.name", "my-cluster-name");
  21. // This instructs the sink to emit after every element, otherwise they would be buffered
  22. config.put("bulk.flush.max.actions", "1");
  24. List<InetSocketAddress> transportAddresses = new ArrayList<>();
  25. transportAddresses.add(new InetSocketAddress(InetAddress.getByName("127.0.0.1"), 9300));
  26. transportAddresses.add(new InetSocketAddress(InetAddress.getByName("10.2.3.1"), 9300));
  28. input.addSink(new ElasticsearchSink<>(config, transportAddresses, new ElasticsearchSinkFunction<String>() {
  29.     public IndexRequest createIndexRequest(String element) {
  30.         Map<String, String> json = new HashMap<>();
  31.         json.put("data", element);
  33.         return Requests.indexRequest()
  34.                 .index("my-index")
  35.                 .type("my-type")
  36.                 .source(json);
  37.     }
  39.     @Override
  40.     public void process(String element, RuntimeContext ctx, RequestIndexer indexer) {
  41.         indexer.add(createIndexRequest(element));
  42.     }
  43. }));
  1. import org.apache.flink.api.common.functions.RuntimeContext;
  2. import org.apache.flink.streaming.api.datastream.DataStream;
  3. import org.apache.flink.streaming.connectors.elasticsearch.ElasticsearchSinkFunction;
  4. import org.apache.flink.streaming.connectors.elasticsearch.RequestIndexer;
  5. import org.apache.flink.streaming.connectors.elasticsearch6.ElasticsearchSink;
  7. import org.apache.http.HttpHost;
  8. import org.elasticsearch.action.index.IndexRequest;
  9. import org.elasticsearch.client.Requests;
  11. import java.util.ArrayList;
  12. import java.util.HashMap;
  13. import java.util.List;
  14. import java.util.Map;
  16. DataStream<String> input = ...;
  18. List<HttpHost> httpHosts = new ArrayList<>();
  19. httpHosts.add(new HttpHost("127.0.0.1", 9200, "http"));
  20. httpHosts.add(new HttpHost("10.2.3.1", 9200, "http"));
  22. // use a ElasticsearchSink.Builder to create an ElasticsearchSink
  23. ElasticsearchSink.Builder<String> esSinkBuilder = new ElasticsearchSink.Builder<>(
  24.     httpHosts,
  25.     new ElasticsearchSinkFunction<String>() {
  26.         public IndexRequest createIndexRequest(String element) {
  27.             Map<String, String> json = new HashMap<>();
  28.             json.put("data", element);
  30.             return Requests.indexRequest()
  31.                     .index("my-index")
  32.                     .type("my-type")
  33.                     .source(json);
  34.         }
  36.         @Override
  37.         public void process(String element, RuntimeContext ctx, RequestIndexer indexer) {
  38.             indexer.add(createIndexRequest(element));
  39.         }
  40.     }
  41. );
  43. // configuration for the bulk requests; this instructs the sink to emit after every element, otherwise they would be buffered
  44. esSinkBuilder.setBulkFlushMaxActions(1);
  46. // provide a RestClientFactory for custom configuration on the internally created REST client
  47. esSinkBuilder.setRestClientFactory(
  48.   restClientBuilder -> {
  49.     restClientBuilder.setDefaultHeaders(...)
  50.     restClientBuilder.setMaxRetryTimeoutMillis(...)
  51.     restClientBuilder.setPathPrefix(...)
  52.     restClientBuilder.setHttpClientConfigCallback(...)
  53.   }
  54. );
  56. // finally, build and add the sink to the job's pipeline
  57. input.addSink(esSinkBuilder.build());
  1. import org.apache.flink.api.common.functions.RuntimeContext
  2. import org.apache.flink.streaming.api.datastream.DataStream
  3. import org.apache.flink.streaming.connectors.elasticsearch.ElasticsearchSinkFunction
  4. import org.apache.flink.streaming.connectors.elasticsearch.RequestIndexer
  5. import org.apache.flink.streaming.connectors.elasticsearch5.ElasticsearchSink
  7. import org.elasticsearch.action.index.IndexRequest
  8. import org.elasticsearch.client.Requests
  10. import java.net.InetAddress
  11. import java.net.InetSocketAddress
  12. import java.util.ArrayList
  13. import java.util.HashMap
  14. import java.util.List
  15. import java.util.Map
  17. val input: DataStream[String] = ...
  19. val config = new java.util.HashMap[String, String]
  20. config.put("cluster.name", "my-cluster-name")
  21. // This instructs the sink to emit after every element, otherwise they would be buffered
  22. config.put("bulk.flush.max.actions", "1")
  24. val transportAddresses = new java.util.ArrayList[InetSocketAddress]
  25. transportAddresses.add(new InetSocketAddress(InetAddress.getByName("127.0.0.1"), 9300))
  26. transportAddresses.add(new InetSocketAddress(InetAddress.getByName("10.2.3.1"), 9300))
  28. input.addSink(new ElasticsearchSink(config, transportAddresses, new ElasticsearchSinkFunction[String] {
  29.   def createIndexRequest(element: String): IndexRequest = {
  30.     val json = new java.util.HashMap[String, String]
  31.     json.put("data", element)
  33.     return Requests.indexRequest()
  34.             .index("my-index")
  35.             .type("my-type")
  36.             .source(json)
  37.   }
  38. }))
  1. import org.apache.flink.api.common.functions.RuntimeContext
  2. import org.apache.flink.streaming.api.datastream.DataStream
  3. import org.apache.flink.streaming.connectors.elasticsearch.ElasticsearchSinkFunction
  4. import org.apache.flink.streaming.connectors.elasticsearch.RequestIndexer
  5. import org.apache.flink.streaming.connectors.elasticsearch6.ElasticsearchSink
  7. import org.apache.http.HttpHost
  8. import org.elasticsearch.action.index.IndexRequest
  9. import org.elasticsearch.client.Requests
  11. import java.util.ArrayList
  12. import java.util.List
  14. val input: DataStream[String] = ...
  16. val httpHosts = new java.util.ArrayList[HttpHost]
  17. httpHosts.add(new HttpHost("127.0.0.1", 9200, "http"))
  18. httpHosts.add(new HttpHost("10.2.3.1", 9200, "http"))
  20. val esSinkBuilder = new ElasticsearchSink.Builder[String](
  21.   httpHosts,
  22.   new ElasticsearchSinkFunction[String] {
  23.      def process(element: String, ctx: RuntimeContext, indexer: RequestIndexer) {
  24.           val json = new java.util.HashMap[String, String]
  25.           json.put("data", element)
  27.           val rqst: IndexRequest = Requests.indexRequest
  28.             .index("my-index")
  29.             .`type`("my-type")
  30.             .source(json)
  32.           indexer.add(rqst)
  33.      } 
  34.   }
  35. )
  37. // configuration for the bulk requests; this instructs the sink to emit after every element, otherwise they would be buffered
  38. esSinkBuilder.setBulkFlushMaxActions(1)
  40. // provide a RestClientFactory for custom configuration on the internally created REST client
  41. esSinkBuilder.setRestClientFactory(new RestClientFactory {
  42.   override def configureRestClientBuilder(restClientBuilder: RestClientBuilder): Unit = {
  43.        restClientBuilder.setDefaultHeaders(...)
  44.        restClientBuilder.setMaxRetryTimeoutMillis(...)
  45.        restClientBuilder.setPathPrefix(...)
  46.        restClientBuilder.setHttpClientConfigCallback(...)
  47.   }
  48. })
  50. // finally, build and add the sink to the job's pipeline
  51. input.addSink(esSinkBuilder.build)

For Elasticsearch versions that still uses the now deprecated TransportClient to communicate with the Elasticsearch cluster (i.e., versions equal or below 5.x), note how a Map of Strings is used to configure the ElasticsearchSink. This config map will be directly forwarded when creating the internally used TransportClient. The configuration keys are documented in the Elasticsearch documentation here. Especially important is the cluster.name parameter that must correspond to the name of your cluster.

For Elasticsearch 6.x and above, internally, the RestHighLevelClient is used for cluster communication. By default, the connector uses the default configurations for the REST client. To have custom configuration for the REST client, users can provide a RestClientFactory implementation when setting up the ElasticsearchClient.Builder that builds the sink.

Also note that the example only demonstrates performing a single index request for each incoming element. Generally, the ElasticsearchSinkFunction can be used to perform multiple requests of different types (ex., DeleteRequest, UpdateRequest, etc.).

Internally, each parallel instance of the Flink Elasticsearch Sink uses a BulkProcessor to send action requests to the cluster. This will buffer elements before sending them in bulk to the cluster. The BulkProcessor executes bulk requests one at a time, i.e. there will be no two concurrent flushes of the buffered actions in progress.

Elasticsearch Sinks and Fault Tolerance

With Flink’s checkpointing enabled, the Flink Elasticsearch Sink guarantees at-least-once delivery of action requests to Elasticsearch clusters. It does so by waiting for all pending action requests in the BulkProcessor at the time of checkpoints. This effectively assures that all requests before the checkpoint was triggered have been successfully acknowledged by Elasticsearch, before proceeding to process more records sent to the sink.

More details on checkpoints and fault tolerance are in the fault tolerance docs.

To use fault tolerant Elasticsearch Sinks, checkpointing of the topology needs to be enabled at the execution environment:

  1. final StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
  2. env.enableCheckpointing(5000); // checkpoint every 5000 msecs
  1. val env = StreamExecutionEnvironment.getExecutionEnvironment()
  2. env.enableCheckpointing(5000) // checkpoint every 5000 msecs

NOTE: Users can disable flushing if they wish to do so, by calling disableFlushOnCheckpoint() on the created ElasticsearchSink. Be aware that this essentially means the sink will not provide any strong delivery guarantees anymore, even with checkpoint for the topology enabled.

Handling Failing Elasticsearch Requests

Elasticsearch action requests may fail due to a variety of reasons, including temporarily saturated node queue capacity or malformed documents to be indexed. The Flink Elasticsearch Sink allows the user to specify how request failures are handled, by simply implementing an ActionRequestFailureHandler and providing it to the constructor.

Below is an example:

  1. DataStream<String> input = ...;
  3. input.addSink(new ElasticsearchSink<>(
  4.     config, transportAddresses,
  5.     new ElasticsearchSinkFunction<String>() {...},
  6.     new ActionRequestFailureHandler() {
  7.         @Override
  8.         void onFailure(ActionRequest action,
  9.                 Throwable failure,
  10.                 int restStatusCode,
  11.                 RequestIndexer indexer) throw Throwable {
  13.             if (ExceptionUtils.findThrowable(failure, EsRejectedExecutionException.class).isPresent()) {
  14.                 // full queue; re-add document for indexing
  15.                 indexer.add(action);
  16.             } else if (ExceptionUtils.findThrowable(failure, ElasticsearchParseException.class).isPresent()) {
  17.                 // malformed document; simply drop request without failing sink
  18.             } else {
  19.                 // for all other failures, fail the sink
  20.                 // here the failure is simply rethrown, but users can also choose to throw custom exceptions
  21.                 throw failure;
  22.             }
  23.         }
  24. }));
  1. val input: DataStream[String] = ...
  3. input.addSink(new ElasticsearchSink(
  4.     config, transportAddresses,
  5.     new ElasticsearchSinkFunction[String] {...},
  6.     new ActionRequestFailureHandler {
  7.         @throws(classOf[Throwable])
  8.         override def onFailure(ActionRequest action,
  9.                 Throwable failure,
  10.                 int restStatusCode,
  11.                 RequestIndexer indexer) {
  13.             if (ExceptionUtils.findThrowable(failure, EsRejectedExecutionException.class).isPresent()) {
  14.                 // full queue; re-add document for indexing
  15.                 indexer.add(action)
  16.             } else if (ExceptionUtils.findThrowable(failure, ElasticsearchParseException.class).isPresent()) {
  17.                 // malformed document; simply drop request without failing sink
  18.             } else {
  19.                 // for all other failures, fail the sink
  20.                 // here the failure is simply rethrown, but users can also choose to throw custom exceptions
  21.                 throw failure
  22.             }
  23.         }
  24. }))

The above example will let the sink re-add requests that failed due to queue capacity saturation and drop requests with malformed documents, without failing the sink. For all other failures, the sink will fail. If a ActionRequestFailureHandler is not provided to the constructor, the sink will fail for any kind of error.

Note that onFailure is called for failures that still occur only after the BulkProcessor internally finishes all backoff retry attempts. By default, the BulkProcessor retries to a maximum of 8 attempts with an exponential backoff. For more information on the behaviour of the internal BulkProcessor and how to configure it, please see the following section.

By default, if a failure handler is not provided, the sink uses a NoOpFailureHandler that simply fails for all kinds of exceptions. The connector also provides a RetryRejectedExecutionFailureHandler implementation that always re-add requests that have failed due to queue capacity saturation.

IMPORTANT: Re-adding requests back to the internal BulkProcessor on failures will lead to longer checkpoints, as the sink will also need to wait for the re-added requests to be flushed when checkpointing. For example, when using RetryRejectedExecutionFailureHandler, checkpoints will need to wait until Elasticsearch node queues have enough capacity for all the pending requests. This also means that if re-added requests never succeed, the checkpoint will never finish.

Configuring the Internal Bulk Processor

The internal BulkProcessor can be further configured for its behaviour on how buffered action requests are flushed, by setting the following values in the provided Map<String, String>:

  • bulk.flush.max.actions: Maximum amount of actions to buffer before flushing.
  • bulk.flush.max.size.mb: Maximum size of data (in megabytes) to buffer before flushing.
  • bulk.flush.interval.ms: Interval at which to flush regardless of the amount or size of buffered actions.

For versions 2.x and above, configuring how temporary request errors are retried is also supported:

  • bulk.flush.backoff.enable: Whether or not to perform retries with backoff delay for a flush if one or more of its actions failed due to a temporary EsRejectedExecutionException.
  • bulk.flush.backoff.type: The type of backoff delay, either CONSTANT or EXPONENTIAL
  • bulk.flush.backoff.delay: The amount of delay for backoff. For constant backoff, this is simply the delay between each retry. For exponential backoff, this is the initial base delay.
  • bulk.flush.backoff.retries: The amount of backoff retries to attempt.

More information about Elasticsearch can be found here.

Packaging the Elasticsearch Connector into an Uber-Jar

For the execution of your Flink program, it is recommended to build a so-called uber-jar (executable jar) containing all your dependencies (see here for further information).

Alternatively, you can put the connector’s jar file into Flink’s lib/ folder to make it available system-wide, i.e. for all job being run.