Kafka event listener#
Примечание
Ниже приведена оригинальная документация Trino. Скоро мы ее переведем на русский язык и дополним полезными примерами.
The Kafka event listener plugin allows streaming of query events to an external Kafka-compatible system. The query history in the Kafka topic can then be accessed directly in Kafka, via Trino in a catalog using the Kafka connector or many downstream systems processing and storing the data.
Rationale#
This event listener is a first step to store the query history of your Trino cluster. The query events can provide CPU and memory usage metrics, what data is being accessed with resolution down to specific columns, and metadata about the query processing.
Running the capture system separate from Trino reduces the performance impact and avoids downtime for non-client-facing changes.
Requirements#
You need to perform the following steps:
Provide an Kafka service that is network-accessible to Trino.
Configure
kafka-event-listener.broker-endpoints
in the event listener properties file with the URI of the serviceConfigure what events to send as detailed in Configuration
Configuration#
To configure the Kafka event listener, create an event listener properties
file in etc
named kafka-event-listener.properties
with the following
contents as an example of a minimal required configuration:
event-listener.name=kafka
kafka-event-listener.broker-endpoints=kafka.example.com:9093
kafka-event-listener.created-event.topic=query_create
kafka-event-listener.completed-event.topic=query_complete
kafka-event-listener.client-id=trino-example
Add etc/kafka-event-listener.properties
to event-listener.config-files
in Config properties:
event-listener.config-files=etc/kafka-event-listener.properties,...
Use the following properties for further configuration.
Property name |
Description |
Default |
---|---|---|
|
Comma-separated list of Kafka broker endpoints with URL and port, for
example |
|
|
Boolean switch to enable anonymization of the event data in Trino before it is sent to Kafka. |
|
|
String identifier for the Trino cluster to allow distinction in Kafka, if multiple Trino clusters send events to the same Kafka system. |
|
|
Boolean switch to control publishing of query creation events. |
|
|
Name of the Kafka topic for the query creation event data. |
|
|
Boolean switch to control publishing of split completion events. |
|
|
Name of the Kafka topic for the split completion event data. |
|
|
Boolean switch to control publishing of query completion events. |
|
|
Name of the Kafka topic for the query completion event data. |
|
|
Comma-separated list of field names to exclude from the Kafka event, for
example |
|
|
Comma-separated list of key-value pairs to specify Kafka client configuration
overrides, for example |
|
|
Timeout duration to complete a Kafka request. Minimum
value of |
|
|
Kafka publisher initialization can fail due to network issues reaching the Kafka brokers. This boolean switch controls whether to throw an exception in such cases. |
|
|
When set, Kafka events are sent with additional metadata populated from
environment variables. For example, if the value is |