nsivabalan commented on code in PR #5269:
URL: https://github.com/apache/hudi/pull/5269#discussion_r846563805
##########
hudi-client/hudi-spark-client/src/main/java/org/apache/hudi/client/HoodieSparkClusteringClient.java:
##########
@@ -53,11 +53,15 @@ public void cluster(HoodieInstant instant) throws
IOException {
SparkRDDWriteClient<T> writeClient = (SparkRDDWriteClient<T>)
clusteringClient;
Option<HoodieCommitMetadata> commitMetadata =
writeClient.cluster(instant.getTimestamp(), true).getCommitMetadata();
Stream<HoodieWriteStat> hoodieWriteStatStream =
commitMetadata.get().getPartitionToWriteStats().entrySet().stream().flatMap(e ->
- e.getValue().stream());
+ e.getValue().stream());
long errorsCount =
hoodieWriteStatStream.mapToLong(HoodieWriteStat::getTotalWriteErrors).sum();
if (errorsCount > 0) {
// TODO: Should we treat this fatal and throw exception?
LOG.error("Clustering for instant (" + instant + ") failed with write
errors");
}
+ if (writeClient != clusteringClient) {
+ // Clustering write client has been updated with new schema, closing the
old one
+ writeClient.close();
Review Comment:
guess, we could miss to close the old write client if there is no clustering
inflight. i.e. when async clustering thread is just waiting and if write client
is updated (due to schema change), we might over ride w/ new write client and
we may not close the old one. Can we ensure we close it out in all cases.
--
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]
