alexey.rassokhin/tet123
1
0
Fork 0
Code Issues Pull Requests 1 Packages Projects Releases Wiki Activity

DBZ-8133 Add async engine options also to Debezium server documentation

Browse Source
This commit is contained in:
Vojtech Juranek 2024-08-16 11:01:17 +02:00 committed by Jiri Pechanec
parent a4ff03716b
commit 583b47dc85
1 changed files with 42 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

42
documentation/modules/ROOT/pages/operations/debezium-server.adoc
View File

@ -478,6 +478,48 @@ The configuration will need to contain the list of predicates, implementation cl
|===
[id="debezium-async-engine-properties"]
=== Asynchronous Engine Properties
By default, Debezium server uses `AsyncEmbeddedEngine` as the processing engine.
You can configure following options for asynchronous embedded engine:
[cols="35%a,10%a,55%a"]
|===
|Property
|Default
|Description
|`record.processing.threads`
|The number of available cores.
|The number of threads to use for processing CDC records.
If you want to use all available cores on given machine, you can use the `AVAILABLE_CORES` placeholder.
If the value is not specified, threads are created as needed, using Java https://docs.oracle.com/en/java/javase/17/docs/api/java.base/java/util/concurrent/Executors.html#newCachedThreadPool()[cached thread pool].
|`record.processing.shutdown.timeout.ms`
|1000
|Maximum time in milliseconds to wait for processing submitted records after a task shutdown is called.
|`record.processing.order`
|`ORDERED`
|Determines how the records should be produced.
`ORDERED`:: Records are processed sequentially; that is, they are produced in the order in which they were obtained from the database.
`UNORDERED`:: Records are processed non-sequentially; that is, they can be produced in an different order than in the source database.
The non-sequential processing of the `UNORDERED` option results in better throughput, because records are produced immediately after any SMT processing and message serialization is complete, without waiting for other records.
This option doesn't have any effect when the `ChangeConsumer` method is provided to the engine.
|`record.processing.with.serial.consumer`
|`false`
|Specifies whether the default `ChangeConsumer` should be created from the provided `Consumer`, resulting in serial `Consumer` processing.
This option has no effect if you specified the `ChangeConsumer` interface when you used the API to create the engine.
|`task.management.timeout.ms`
|180,000 (3 min)
|Time, in milliseconds, that the engine waits for a task's lifecycle management operations (starting and stopping) to complete.
|===
[id="debezium-additional-configuration-options"]
=== Additional configuration