[
https://issues.apache.org/jira/browse/DRILL-1197?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14078383#comment-14078383
]
Xiao Meng commented on DRILL-1197:
----------------------------------
Hi Parth,
I agree. It is better to separate a JIRA [#DRILL-1219]. I cannot reproduce the
problem (handleHandleshakeTimeout trigger HandleRead and signal errors) at the
moment and previously the application seems confused and cannot identify the
right error type.
{quote}
Also, after shutdown, is the client's connection still usable? I don't
think it is. If it is not, then we have a much bigger problem. We cannot
allow a client connection to die because the server took too long to
respond!
{quote}
If we use shutdown_both, the socket is not usable anymore. In our application,
we will open a new connection to re-sumbit the query, so it is not a problem.
If I understand it correctly, you assume that we should reuse the socket
connection after timed out?
> C++ Client. Differentiate socket/handshake/query timeout for deadline timer.
> -----------------------------------------------------------------------------
>
> Key: DRILL-1197
> URL: https://issues.apache.org/jira/browse/DRILL-1197
> Project: Apache Drill
> Issue Type: New Feature
> Components: Client - C++
> Reporter: Xiao Meng
> Assignee: Xiao Meng
> Fix For: 0.5.0
>
> Attachments: DRILL-1197.1.patch.txt
>
>
> [#DRILL-852] introduced deadline timer to C++ Client. It use a global setting
> of timeout for all reading operation. We should differentiate timeout for
> different operations and also populate the status to client library user.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
