danny0405 commented on code in PR #19041:
URL: https://github.com/apache/hudi/pull/19041#discussion_r3446262106
##########
hudi-common/src/main/java/org/apache/hudi/common/util/CleanerUtils.java:
##########
@@ -151,8 +152,16 @@ public static Option<HoodieInstant>
getEarliestCommitToRetain(
} else if (cleaningPolicy == HoodieCleaningPolicy.KEEP_LATEST_BY_HOURS) {
ZonedDateTime latestDateTime = ZonedDateTime.ofInstant(latestInstant,
timeZone.getZoneId());
String earliestTimeToRetain =
TimelineUtils.formatDate(Date.from(latestDateTime.minusHours(hoursRetained).toInstant()));
- earliestCommitToRetain =
Option.fromJavaOptional(completedCommitsTimeline.getInstantsAsStream().filter(i
-> compareTimestamps(i.requestedTime(),
- GREATER_THAN_OR_EQUALS, earliestTimeToRetain)).findFirst());
+ earliestCommitToRetain =
Option.fromJavaOptional(completedCommitsTimeline.getInstantsAsStream()
+ .filter(i -> compareTimestamps(i.requestedTime(),
GREATER_THAN_OR_EQUALS, earliestTimeToRetain))
+ .findFirst());
+
+ Option<HoodieInstant> earliestPendingCommit = commitsTimeline.filter(s
-> !s.isCompleted()).firstInstant();
+ if (earliestPendingCommit.isPresent()
+ && (!earliestCommitToRetain.isPresent()
+ || compareTimestamps(earliestPendingCommit.get().requestedTime(),
LESSER_THAN, earliestCommitToRetain.get().requestedTime()))) {
+ earliestCommitToRetain =
completedCommitsTimeline.findInstantsBefore(earliestPendingCommit.get().requestedTime()).lastInstant();
Review Comment:
> Yes, findInstantsBefore(...).lastInstant() can return empty in that case.
Downstream, an empty ECTR means no partition scan/no cleaning for commit/hour
based cleanin
Could you elaborate why the cleaing be blocked by a pending instant here?
since we have completion time file slicing(pending files always belong to the
latest one), it is okey to forward the cleaning, and the ECTR should be based
on completion time too?
--
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]
