[
https://issues.apache.org/jira/browse/FLINK-7416?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16212725#comment-16212725
]
ASF GitHub Bot commented on FLINK-7416:
---------------------------------------
Github user pnowojski commented on a diff in the pull request:
https://github.com/apache/flink/pull/4533#discussion_r145980535
--- Diff:
flink-runtime/src/test/java/org/apache/flink/runtime/io/network/netty/PartitionRequestClientHandlerTest.java
---
@@ -208,6 +211,53 @@ public void testCancelBeforeActive() throws Exception {
client.cancelRequestFor(inputChannel.getInputChannelId());
}
+ /**
+ * Verifies that {@link RemoteInputChannel} is enqueued in the
pipeline, and
+ * {@link AddCredit} message is sent to the producer.
+ */
+ @Test
+ public void testNotifyCreditAvailable() throws Exception {
+ final CreditBasedClientHandler handler = new
CreditBasedClientHandler();
+ final EmbeddedChannel channel = new EmbeddedChannel(handler);
+
+ final RemoteInputChannel inputChannel =
mock(RemoteInputChannel.class);
--- End diff --
I know that mocking was extensively used so far, but we had quite bad
experience with it.
I don't agree that it is not interacting, because it is for example
checking `inputChannel.isReleased()`, which you need to mock one test below.
Now imagine that during some refactor/implementing new feature someone modifies
handlers code, by adding some additional interaction/sanity check on
`inputChannel`. With mocks it will lead with thousands of changes in all of the
mocked unit tests.
> Implement Netty receiver outgoing pipeline for credit-based
> -----------------------------------------------------------
>
> Key: FLINK-7416
> URL: https://issues.apache.org/jira/browse/FLINK-7416
> Project: Flink
> Issue Type: Sub-task
> Components: Network
> Reporter: zhijiang
> Assignee: zhijiang
> Fix For: 1.4.0
>
>
> This is a part of work for credit-based network flow control.
> The related works are :
> * We define a new message called {{AddCredit}} to notify the incremental
> credit during data shuffle.
> * Whenever an {{InputChannel}}’s unannounced credit goes up from zero, the
> channel is enqueued in the pipeline.
> * Whenever the channel becomes writable, it takes the next {{InputChannel}}
> and sends its unannounced credit. The credit is reset to zero after each sent.
> * That way, messages are sent as often as the network has capacity and
> contain as much credit as available for the channel at that point in time.
> Otherwise, it would only add latency to the announcements and not increase
> throughput.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
