[
https://issues.apache.org/jira/browse/GEODE-3563?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16472462#comment-16472462
]
Bruce Schuchardt commented on GEODE-3563:
-----------------------------------------
There was a problem with the fix for this ticket. The changes caused all
sockets, ssl & non-ssl, to have timeouts established. This was causing idle
TCPConduit connections to time out after a minute, whereas before they did not
timeout.
I'm modifying the fix to establish timeouts in the callers of the method and am
changing the method itself to override the timeout for the duration of the
handshake and then restore it to its old value.
> SSL socket handling problems in TCPConduit run
> ----------------------------------------------
>
> Key: GEODE-3563
> URL: https://issues.apache.org/jira/browse/GEODE-3563
> Project: Geode
> Issue Type: Bug
> Components: messaging
> Reporter: Vahram Aharonyan
> Assignee: Galen O'Sullivan
> Priority: Critical
> Labels: pull-request-available
> Fix For: 1.6.0
>
> Time Spent: 1h 20m
> Remaining Estimate: 0h
>
> Here are two cases that seems to problematic in TCPConduit.run flow:
> 1. TCPConduit.run() has no action performed for the case when SSLException is
> thrown from sslSocket.startHandshake(), as a result the socket remains open.
> Catch block from the end of configureServerSSLSocket() will just report a
> fatal error(even it seem that this portion is going to be removed in 1.2.1
> according to GEODE-3393) and re-throw the exception.
> 2. configureServerSSLSocket call is performed without setting socket timeout
> before that. This can bring to run thread blocking case if read initiated
> from the SSL handshake flow will not return. Linking to similar issues
> observed with other acceptors previously: GEODE-2898, GEODE-3023.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
