yux created FLINK-37278:
---------------------------
Summary: Regular schema evolution process is too slow when
starting from specific offset
Key: FLINK-37278
URL: https://issues.apache.org/jira/browse/FLINK-37278
Project: Flink
Issue Type: Bug
Components: Flink CDC
Affects Versions: cdc-3.3.0
Reporter: yux
Currently, regular schema evolution process is too slow if:
# Uses MySQL as source
# Starts from specific binlog offset
# Captures huge amount of tables in capture list
Each CreateTableEvent will take at least 1 second to finish, and long time
backpressure would cause MySQL binlog client to timeout and job will fail then.
---
We could optimize the schema evolution performance by replacing client-side
loop-querying schema change request with `CompletableFuture`s (just like we did
in distributed schema evolution topology), which allows the SchemaCoordinator
to decide when to accept a schema change requests, needless to wait for
clients' spontaneous request.
---
Originally reported by [~hiliuxg].
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
