github-actions[bot] commented on code in PR #65588:
URL: https://github.com/apache/doris/pull/65588#discussion_r3593746028
##########
fs_brokers/cdc_client/src/main/java/io/debezium/connector/mysql/MySqlStreamingChangeEventSource.java:
##########
@@ -408,7 +416,10 @@ protected void handleEvent(
eventDispatcher.dispatchHeartbeatEvent(partition, offsetContext);
// Capture that we've completed another event ...
- offsetContext.completeEvent();
+ // OceanBase heartbeat events must not participate in restart
event counting.
+ if (shouldCompleteEvent(excludeHeartbeatFromEventCount,
eventType)) {
Review Comment:
[P1] Apply heartbeat exclusion to the replay skip count too
`initialEventsToSkip` is restored from the events counted by
`completeEvent()`, but this change filters OceanBase heartbeats only when
incrementing that count. While replaying a partial transaction, lines 427-431
still decrement the count for such a heartbeat. For example, with a saved
prefix of BEGIN plus one row, replayed BEGIN leaves one skip; a
timing-dependent HEARTBEAT consumes it, clears `skipEvent`, and the
already-processed row is emitted again. Please use the same
`shouldCompleteEvent(...)` decision for both counting and replay decrement, and
add a deterministic BEGIN/HEARTBEAT/row restart test.
##########
regression-test/data/job_p0/streaming_job/cdc/test_streaming_oceanbase_job_all_type.out:
##########
@@ -0,0 +1,8 @@
+-- This file is automatically generated. You should know what you did if you
want to edit this
+-- !oceanbase_all_type_snapshot --
+1 200 18446744073709551610 123456789.12345 true 2026-07-10
2026-07-10T10:11:12.123456 2026-07-10T10:11:12.123456 abc
snapshot snapshot text 41514944 {"id": 1, "name": "snapshot"}
A,C NEW 4367734D
+2 \N \N \N \N \N \N \N \N \N
\N \N \N \N \N \N
+
+-- !oceanbase_all_type_incremental --
+1 200 18446744073709551610 1.25000 true 2026-07-10
2026-07-10T10:11:12.123456 2026-07-10T02:11:12.123456 abc updated
snapshot text 41514944 {"id":1,"name":"snapshot"} A,C NEW
4367734D
Review Comment:
[P1] Keep OceanBase TIMESTAMP values consistent across snapshot and CDC
This expected result records `id=1.event_timestamp` as `10:11:12.123456`
after the snapshot, but as `02:11:12.123456` after the UPDATE at suite lines
123-124, which changes only `amount` and `variable_text`. With
`serverTimezone=UTC`, the established MySQL-compatible contract keeps an
unchanged TIMESTAMP at the same instant across snapshot and binlog paths; here
the first unrelated update silently shifts it by eight hours. Please align
OceanBase snapshot and binlog timestamp decoding, keep this untouched value
stable, and regenerate the result with an explicit regression assertion for the
transition.
##########
fe/fe-core/src/main/java/org/apache/doris/job/common/DataSourceType.java:
##########
@@ -19,5 +19,6 @@
public enum DataSourceType {
MYSQL,
- POSTGRES
+ POSTGRES,
+ OCEANBASE
Review Comment:
[P2] Wire OCEANBASE through the cdc_stream entry path
Adding this shared enum value makes `cdc_stream(type="oceanbase", ...)` pass
`DataSourceType.valueOf()`, but `CdcStreamTableValuedFunction.validate()` still
handles only MYSQL and POSTGRES and then throws `Unsupported type: OCEANBASE`.
That blocks both standalone TVF use and streaming INSERT jobs before they can
reach the already-compatible TVF offset provider, `getRemoteDbName()`, or the
new CDC reader. Please add the MySQL-compatible OceanBase validation/preflight
here and cover both cdc_stream forms.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
