pnowojski commented on a change in pull request #17238:
URL: https://github.com/apache/flink/pull/17238#discussion_r710779840
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/netty/PartitionRequestServerHandler.java
##########
@@ -82,16 +82,17 @@ protected void channelRead0(ChannelHandlerContext ctx,
NettyMessage msg) throws
LOG.debug("Read channel on {}: {}.",
ctx.channel().localAddress(), request);
+ // Always register reader before requesting the subpartition
in order to
+ // subsequent requests will be sure that PartitionRequest was
received already
+ // even if subpartition view has not created yet.
+ NetworkSequenceViewReader reader =
+ outboundQueue.notifyReaderCreated(
+ new CreditBasedSequenceNumberingViewReader(
+ request.receiverId, request.credit,
outboundQueue));
Review comment:
> until restart happens
I'm not sure, but it might be until TM restarts, which might be never, while
number of those zombie readers would be increasing with the number of scheduled
jobs in the cluster. Think about a use case where there is a single cluster and
many adhoc short lived jobs.
--
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]
