wang-haihua commented on code in PR #3805:
URL: https://github.com/apache/celeborn/pull/3805#discussion_r3796461336
##########
client/src/main/java/org/apache/celeborn/client/ShuffleClientImpl.java:
##########
@@ -2048,6 +2048,19 @@ public CelebornInputStream readPartition(
logger.warn("Shuffle data is empty for shuffle {} partition {}.",
shuffleId, partitionId);
return CelebornInputStream.empty();
} else {
+ if (logger.isInfoEnabled()) {
+ Set<String> workerHosts = new LinkedHashSet<>();
+ for (PartitionLocation location : locations) {
+ if (location != null) {
+ workerHosts.add(location.getHost());
+ }
+ }
+ logger.info(
+ "Reduce task for shuffle {} partition {} reads data from worker
hosts: {}",
Review Comment:
Yes, it does address a similar problem to CELEBORN-2208. The difference is
that CELEBORN-2208 is an anomaly signal — in my case I'm on 0.6.3, but the log
never printed even though Shuffle Read Blocked Time was 14–19 min. This PR
always gives that mapping, so a slow task in the Spark UI can be
cross-referenced to the workers it read from, which helps with troubleshooting.
--
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]
