[
https://issues.apache.org/jira/browse/HADOOP-12605?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15056880#comment-15056880
]
Xiao Chen commented on HADOOP-12605:
------------------------------------
Hi [~iwasakims],
Thanks for the explanation and the patch. Sorry for my delayed response.
I'm not sure about relaxing the check between
{{server.getNumOpenConnections()}} and {{maxAccept}} - that would no longer
check the intention of 'we didn't go over'. I hope the original author [~daryn]
could take a look and advice.
Also, the failure happens for {{checkBlocking(1, 1, 1)}}, so I'll discuss this
case here for simplicity. With the {{firstCallLatch}} in place in the test, I
didn't find anywhere the server can accept more connections. Could you further
explain your theory?
>From my understanding, the call goes briefly like this: client call ->
>server's listener -> add to readerQ -> reader take from readerQ -> put to
>callQ -> handler take from callQ -> call (blocked by the latch). In the (1, 1,
>1) case we only have 1 reader with readerQ size=1, and callQ size =1, so IMO
>neither the round robin nor the loop of taking pendingConnections should
>contribute anything.
I'm trying to hack the test to result in a similar failure but haven't
succeeded so far. [~iwasakims], would you please share some reproduce code if
possible? Thanks very much.
> Fix intermittent failure of TestIPC.testIpcWithReaderQueuing
> ------------------------------------------------------------
>
> Key: HADOOP-12605
> URL: https://issues.apache.org/jira/browse/HADOOP-12605
> Project: Hadoop Common
> Issue Type: Bug
> Components: test
> Reporter: Masatake Iwasaki
> Assignee: Masatake Iwasaki
> Priority: Minor
> Attachments: HADOOP-12605.001.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
