kbuci commented on code in PR #18302:
URL: https://github.com/apache/hudi/pull/18302#discussion_r2963256186
##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/BaseHoodieTableServiceClient.java:
##########
@@ -1155,13 +1178,37 @@ private List<String>
getInstantsToRollbackForLazyCleanPolicy(HoodieTableMetaClie
if (!expiredInstants.isEmpty()) {
// Only return instants that haven't been completed by other writers
metaClient.reloadActiveTimeline();
- HoodieTimeline refreshedInflightTimeline =
getInflightTimelineExcludeCompactionAndClustering(metaClient);
- return
expiredInstants.stream().filter(refreshedInflightTimeline::containsInstant).collect(Collectors.toList());
+ HoodieTimeline refreshedIncompleteTimeline =
metaClient.getActiveTimeline().filterInflightsAndRequested();
+ return expiredInstants.stream().filter(instantTime ->
+ refreshedIncompleteTimeline.containsInstant(instantTime)
+ ).collect(Collectors.toList());
} else {
return Collections.emptyList();
}
}
+ public boolean isClusteringInstantEligibleForRollback(HoodieTableMetaClient
metaClient, HoodieInstant instant) {
+ return config.isExpirationOfClusteringEnabled()
+ && hasInstantExpired(metaClient, instant.requestedTime(),
config.getClusteringExpirationTimeMins())
+ && ClusteringUtils.isClusteringInstant(
+ metaClient.getActiveTimeline(), instant,
metaClient.getInstantGenerator());
+ }
+
+ private static boolean hasInstantExpired(HoodieTableMetaClient metaClient,
String instantTime, long expirationMins) {
Review Comment:
I wanted to avoid redundant heartbeat checks which is why I didn't do that
here - https://github.com/apache/hudi/pull/18302/changes#r2943866147 ..
Especially since the caller anyway needs to reload the timeline after to
double-check that an inflight instant with expired heartbeat wasn't actually
just a completed instant.
But since this is not intuitive when reading the code, I ended up abandoning
this approach. And now this API also checks heartbeat status.
--
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]
