QiLuo-BD commented on a change in pull request #7186: [FLINK-10941] Keep slots
which contain unconsumed result partitions
URL: https://github.com/apache/flink/pull/7186#discussion_r238217342
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/netty/PartitionRequestQueue.java
##########
@@ -134,10 +138,17 @@ public void cancel(InputChannelID receiverId) {
ctx.pipeline().fireUserEventTriggered(receiverId);
}
- public void close() {
+ public void close() throws IOException {
if (ctx != null) {
ctx.channel().close();
}
+
+ LOG.info("Close all {} readers pending for close.",
readersToClose.size());
Review comment:
If the partition is released once it's transported, then the TM containing
the partition will be released. It will cause a RemoteTransportException
"Connection unexpectedly closed by remote task manager ..." in
CreditBasedPartitionRequestClientHandler's channelInactive() method, as the
Netty connection is shared between TMs.
So I changed this to let writer to close the readers when closing the Netty
connection. This is the smallest possible change I can find. Do you have any
suggestions how we can handle this case better?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
