[
https://issues.apache.org/jira/browse/KAFKA-1400?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14337608#comment-14337608
]
Gwen Shapira commented on KAFKA-1400:
-------------------------------------
Kinda strange, the TCPDUMP looks the same whether I get an exception or not:
{code}
16:52:17.249207 IP localhost.55538 > localhost.9095: Flags [S], seq 3746193634,
win 65535, options [mss 16344,nop,wscale 5,nop,nop,TS val 1314712608 ecr
0,sackOK,eol], length 0
16:52:17.249257 IP localhost.9095 > localhost.55538: Flags [S.], seq 999508519,
ack 3746193635, win 65535, options [mss 16344,nop,wscale 3,nop,nop,TS val
1314712608 ecr 1314712608,sackOK,eol], length 0
16:52:17.249264 IP localhost.55538 > localhost.9095: Flags [.], ack 1, win
12759, options [nop,nop,TS val 1314712608 ecr 1314712608], length 0
16:52:17.249270 IP localhost.9095 > localhost.55538: Flags [.], ack 1, win
38788, options [nop,nop,TS val 1314712608 ecr 1314712608], length 0
16:52:22.436127 IP localhost.55538 > localhost.9095: Flags [P.], seq 1:2, ack
1, win 12759, options [nop,nop,TS val 1314717768 ecr 1314712608], length 1
16:52:22.436162 IP localhost.9095 > localhost.55538: Flags [.], ack 2, win
38788, options [nop,nop,TS val 1314717768 ecr 1314717768], length 0
16:52:22.436169 IP localhost.55538 > localhost.9095: Flags [P.], seq 2:7, ack
1, win 12759, options [nop,nop,TS val 1314717768 ecr 1314717768], length 5
16:52:22.436173 IP localhost.9095 > localhost.55538: Flags [.], ack 7, win
38787, options [nop,nop,TS val 1314717768 ecr 1314717768], length 0
16:52:22.436177 IP localhost.55538 > localhost.9095: Flags [P.], seq 7:47, ack
1, win 12759, options [nop,nop,TS val 1314717768 ecr 1314717768], length 40
16:52:22.436180 IP localhost.9095 > localhost.55538: Flags [.], ack 47, win
38782, options [nop,nop,TS val 1314717768 ecr 1314717768], length 0
16:52:22.484320 IP localhost.9095 > localhost.55538: Flags [P.], seq 1:23, ack
47, win 38782, options [nop,nop,TS val 1314717808 ecr 1314717768], length 22
16:52:22.484353 IP localhost.55538 > localhost.9095: Flags [.], ack 23, win
12758, options [nop,nop,TS val 1314717808 ecr 1314717808], length 0
16:52:22.552365 IP localhost.9095 > localhost.55538: Flags [F.], seq 23, ack
47, win 38782, options [nop,nop,TS val 1314717872 ecr 1314717808], length 0
16:52:22.552372 IP localhost.55538 > localhost.9095: Flags [.], ack 24, win
12758, options [nop,nop,TS val 1314717872 ecr 1314717872], length 0
16:52:22.552376 IP localhost.9095 > localhost.55538: Flags [.], ack 47, win
38782, options [nop,nop,TS val 1314717872 ecr 1314717872], length 0
16:52:28.632196 IP localhost.55538 > localhost.9095: Flags [P.], seq 47:48, ack
24, win 12758, options [nop,nop,TS val 1314723916 ecr 1314717872], length 1
16:52:28.632235 IP localhost.9095 > localhost.55538: Flags [R], seq 999508543,
win 0, length 0
{code}
I get a FIN, I ack it. I send one extra packet and I get a RST. Pretty much as
expected.
Whether I get an exception or not, OTOH is completely random. I'm guessing this
is because the .write() and .flush() methods are async.
As documented for output stream: The <code>flush</code> method of
<code>OutputStream</code> does nothing.
I *think* that if I force a sleep after some bytes were written, I'll always
get an exception. I'm not too happy with this approach.
Any thoughts?
> transient unit test failure in SocketServerTest
> -----------------------------------------------
>
> Key: KAFKA-1400
> URL: https://issues.apache.org/jira/browse/KAFKA-1400
> Project: Kafka
> Issue Type: Bug
> Components: core
> Affects Versions: 0.8.2.0
> Reporter: Jun Rao
> Assignee: Neha Narkhede
> Fix For: 0.8.2.0
>
> Attachments: KAFKA-1400.patch
>
>
> Saw the following transient failure.
> kafka.network.SocketServerTest > testSocketsCloseOnShutdown FAILED
> java.lang.AssertionError: Expected exception: java.net.SocketException
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
