nsivabalan commented on code in PR #7826:
URL: https://github.com/apache/hudi/pull/7826#discussion_r1101898172
##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/client/BaseHoodieTableServiceClient.java:
##########
@@ -707,20 +709,33 @@ protected List<String>
getInstantsToRollback(HoodieTableMetaClient metaClient, H
}
}).collect(Collectors.toList());
} else if (cleaningPolicy.isLazy()) {
- return inflightInstantsStream.filter(instant -> {
- try {
- return heartbeatClient.isHeartbeatExpired(instant.getTimestamp());
- } catch (IOException io) {
- throw new HoodieException("Failed to check heartbeat for instant " +
instant, io);
- }
- }).map(HoodieInstant::getTimestamp).collect(Collectors.toList());
+ return getInstantsToRollbackForLazyCleanPolicy(metaClient,
inflightInstantsStream);
} else if (cleaningPolicy.isNever()) {
return Collections.emptyList();
} else {
throw new IllegalArgumentException("Invalid Failed Writes Cleaning
Policy " + config.getFailedWritesCleanPolicy());
}
}
+ @VisibleForTesting
+ public List<String>
getInstantsToRollbackForLazyCleanPolicy(HoodieTableMetaClient metaClient,
+
Stream<HoodieInstant> inflightInstantsStream) {
+ // Get expired instants, must store them into list before double-checking
+ List<String> expiredInstants = inflightInstantsStream.filter(instant -> {
+ try {
+ // An instant transformed from inflight to completed have no heartbeat
file and will be detected as expired instant here
+ return heartbeatClient.isHeartbeatExpired(instant.getTimestamp());
+ } catch (IOException io) {
+ throw new HoodieException("Failed to check heartbeat for instant " +
instant, io);
+ }
+ }).map(HoodieInstant::getTimestamp).collect(Collectors.toList());
+
+ // Double check whether the heartbeat-expired instant is an inflight
instant
+ metaClient.reloadActiveTimeline();
Review Comment:
active timeline reload is an expensive operation. we should try to minimize.
can you check if exporedInstants is not empty and reload only during such
cases ?
if infact you can return right away if its empty.
--
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]
