zhijiangW commented on a change in pull request #8242: [FLINK-6227][network]
Introduce the PartitionException for downstream task failure
URL: https://github.com/apache/flink/pull/8242#discussion_r286061517
##########
File path:
flink-runtime/src/test/java/org/apache/flink/runtime/io/network/netty/ClientTransportErrorHandlingTest.java
##########
@@ -134,12 +133,10 @@ public Void answer(InvocationOnMock invocation) throws
Throwable {
}).when(rich[1]).onError(isA(LocalTransportException.class));
// First request is successful
- ChannelFuture f = requestClient.requestSubpartition(new
ResultPartitionID(), 0, rich[0], 0);
- assertTrue(f.await().isSuccess());
+ requestClient.requestSubpartition(new ResultPartitionID(), 0,
rich[0], 0);
// Second request is *not* successful
- f = requestClient.requestSubpartition(new ResultPartitionID(),
0, rich[1], 0);
- assertFalse(f.await().isSuccess());
Review comment:
The `rich[1],onError` is actually verified above, then it would call
`sync.countDown()`. Otherwise the test would fail in timeout.
In `PartitionRequestClient#requestSubpartition`, the `delayMs` is 0 then it
would not schedule `tcpChannel.writeAndFlush(request)`. So I thought it wrote
the request in the same thread before. After double checking this, the write is
actually done in netty thread, so I would consider the race conditions here.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
