[
https://issues.apache.org/jira/browse/ZOOKEEPER-2069?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14195173#comment-14195173
]
Hongchao Deng commented on ZOOKEEPER-2069:
------------------------------------------
Hi [~fpj],
Experimented a few cases over past days.
My finding regarding your comments:
bq. Shouldn't isConnected return true in the case channel isn't null instead of
channelFactory? In fact, it sounds like this implementation is instantiating a
new ChannelFactory every time it tries to connect (modulo an attempt already
being in progress). It doesn't sound necessary.
Yes. isConnected had better be based on (channel != null &&
channel.isConnected).
However, I found a problem reusing channelFactory -- Some un-handled errors
that happened in Netty channel closing will propodate in the executors thread
pool and crash later connection tries.
It can be resolved by creating a new channelFactory for every new connect. I
couldn't come up with perfect solutions to isolate channel errors on top of my
head.
> Netty Support for ClientCnxnSocket
> ----------------------------------
>
> Key: ZOOKEEPER-2069
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2069
> Project: ZooKeeper
> Issue Type: Sub-task
> Reporter: Hongchao Deng
> Assignee: Hongchao Deng
> Attachments: QA-run-nettyclient-for-test.patch,
> ZOOKEEPER-2069-v2.patch, ZOOKEEPER-2069-v3.patch, ZOOKEEPER-2069-v4.patch,
> ZOOKEEPER-2069-v5.patch, ZOOKEEPER-2069-v6.patch, ZOOKEEPER-2069.patch,
> draft.patch
>
>
> Review Board: https://reviews.apache.org/r/27244/diff/#
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
