Skip to main content

Metadata Change Event Consumer Job

The Metadata Change Event Consumer is a Spring job which can be deployed by itself, or as part of the Metadata Service.

Its main function is to listen to change proposal events emitted by clients of DataHub which request changes to the Metadata Graph. It then applies these requests against DataHub's storage layer: the Metadata Service.

Today the job consumes from two topics:

  1. MetadataChangeProposal_v1
  2. (Deprecated) MetadataChangeEvent_v4

and produces to the following topics

  1. FailedMetadataChangeProposal_v1
  2. (Deprecated) FailedMetadataChangeEvent_v4

Where does the misleading name Metadata Change Event come from? Well, history. Previously, this job consumed a single MetadataChangeEvent topic which has been deprecated and replaced by per-aspect Metadata Change Proposals. Hence, the name!

Pre-requisites​

  • You need to have JDK8 installed on your machine to be able to build DataHub Metadata Service.

Build​

Metadata Change Event Consumer Job is already built as part of top level build:

./gradlew build

However, if you only want to build Metadata Change Event Consumer Job specifically:

./gradlew :metadata-jobs:mce-consumer-job:build

Dependencies​

Before starting Metadata Change Event Consumer Job, you need to make sure that Kafka, Schema Registry & Zookeeper and DataHub GMS Docker containers are up and running.

Start via Docker image​

Quickest way to try out Metadata Change Event Consumer Job is running the Docker image.

Start via command line​

If you do modify things and want to try it out quickly without building the Docker image, you can also run the application directly from command line after a successful build:

MCP_CONSUMER_ENABLED=true ./gradlew :metadata-jobs:mce-consumer-job:bootRun

Debugging​

To debug with an IDE (i.e. IntelliJ), run the bootRun task with the --debug-jvm flag. This will launch the app and listen on port 5005 for a remote debugger.

MCP_CONSUMER_ENABLED=true ./gradlew :metadata-jobs:mce-consumer-job:bootRun --debug-jvm

Endpoints​

Spring boot actuator has been enabled for MCE Application. healthcheck, metrics and info web endpoints are enabled by default.

healthcheck - http://localhost:9090/actuator/health

metrics - http://localhost:9090/actuator/metrics

To retrieve a specific metric - http://localhost:9090/actuator/metrics/kafka.consumer.records.consumed.total

Standalone MCE Consumer Mode​

In standalone consumer mode, the number of CPU cores + 1 determines the parallelization of the MCE consumer's internal processing. This determines the number of Kafka concurrent consumers, the number of threads available to the local Restli service, and the maximum number SQL connections used by the MCE consumer container.

Note that the effective throughput is limited to the number of partitions configured for the source Kafka topics. Allocating additional consumers, or CPU cores, beyond the number of topic partitions to the MCE standalone consumer(s) will not increase ingestion performance.

Additional Endpoints​

/* - Restli service endpoints (standalone consumer mode)

The Restli service endpoints are exposed for use locally (not via a K8 service) by the MCE consumer job itself. This is only true in standalone mode. When run within GMS, the Restli endpoints are already available and accessed via the K8 service.