hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vihang Karajgaonkar (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-18964) HiveServer2 should not log errors when clients issue connection reset
Date Fri, 13 Apr 2018 18:47:00 GMT

    [ https://issues.apache.org/jira/browse/HIVE-18964?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16437738#comment-16437738
] 

Vihang Karajgaonkar commented on HIVE-18964:
--------------------------------------------

Turns out we cannot fix this in Hive side because {{TTransportFactory.getTransport}} does
not throw a checked exception and hence under-lying layers need to wrap the {{TTransportException}}
in a {{RuntimeException}} which is unnecessarily logged in {{TThreadpoolServer}}. This issue
needs to be fixed on Thrift side only. I spent a lot time to see if we can configure HAProxy
such that it issues a EOF during health-check but I don't think there is a way to configure
that. HAproxy issues a socket reset once the tcp transport is opened which causes a TTransportException
in SASL negotiation phase.

> HiveServer2 should not log errors when clients issue connection reset
> ---------------------------------------------------------------------
>
>                 Key: HIVE-18964
>                 URL: https://issues.apache.org/jira/browse/HIVE-18964
>             Project: Hive
>          Issue Type: Improvement
>          Components: HiveServer2
>    Affects Versions: 3.0.0, 2.3.2
>            Reporter: Vihang Karajgaonkar
>            Assignee: Vihang Karajgaonkar
>            Priority: Minor
>
> HiveServer2 logs an ugly exception trace when clients issue a connection reset. There
is nothing we can do when the connection is reset. This exception trace should either be ignored
or logged only in debug mode.
> Things become worse if you are using load balancer like HAProxy which has its own health
checks. HAProxy issues a connection reset to "quickly" close the connection once it finds
that HS2 is up and available. This spams the logs at a very high frequency and makes them
unusable for debugging purposes.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message