[
https://issues.apache.org/jira/browse/HIVE-5799?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13827377#comment-13827377
]
Navis commented on HIVE-5799:
-----------------------------
[~thejas] Thinking of what you said (ping from client to server), I've realized
I've once failed that a few months ago. I remember thrift does not allow
asynchronous call, which means it allows only one call on wire per connection.
If we wrap and track all the onvocations on connection/statements/resultsets
then we can confirm there is no on-going call and send ping message safely (and
block other calls till ping back is arrived) but it seemed not that easy. Any
better idea?
> session/operation timeout for hiveserver2
> -----------------------------------------
>
> Key: HIVE-5799
> URL: https://issues.apache.org/jira/browse/HIVE-5799
> Project: Hive
> Issue Type: Improvement
> Components: HiveServer2
> Reporter: Navis
> Assignee: Navis
> Priority: Minor
> Attachments: HIVE-5799.1.patch.txt, HIVE-5799.2.patch.txt,
> HIVE-5799.3.patch.txt
>
>
> Need some timeout facility for preventing resource leakages from instable or
> bad clients.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
