wombatu-kun commented on code in PR #20276:
URL: https://github.com/apache/hudi/pull/20276#discussion_r4279968425
##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/BaseHoodieWriteClient.java:
##########
@@ -1710,11 +1712,31 @@ protected void tryUpgrade(HoodieTableMetaClient
metaClient, Option<String> insta
metaClient.reloadTableConfig();
metaClient.reloadActiveTimeline();
}
+
+ backfillTableConfigs(metaClient);
Review Comment:
On Flink, `StreamWriteOperatorCoordinator.start` reaches this backfill
through `HoodieFlinkWriteClient.tryUpgrade` with no lock held, so in a
multi-writer setup its read-modify-write of hoodie.properties can race another
writer's locked update of the same file. Could the backfill run only from the
locked `startCommit` and `doInitTable` paths, given Flink's `startCommit`
already reaches it under the lock?
--
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]
