[
https://issues.apache.org/jira/browse/GEODE-6928?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16876501#comment-16876501
]
ASF subversion and git services commented on GEODE-6928:
--------------------------------------------------------
Commit 48030961f82f91360d8c88f19516758ba0d6affe in geode's branch
refs/heads/feature/GEODE-6930 from Bruce Schuchardt
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=4803096 ]
GEODE-6928 peer-to-peer SSL stream corruption with conserve-sockets=false
Modified the NioSslEngine to not modify the decrypted SSL buffer after
reading a direct-ack response. This allows the readFully method to
correctly see what bytes have already been consumed and correctly
compact the buffer for subsequent reads, if necessary.
The cluster communication test is modified to check for aborted
connections created (retries) during operation distribution. Without the
fix for the problem this check would fail.
> peer-to-peer SSL stream corruption with conserve-sockets=false
> --------------------------------------------------------------
>
> Key: GEODE-6928
> URL: https://issues.apache.org/jira/browse/GEODE-6928
> Project: Geode
> Issue Type: Bug
> Components: messaging
> Reporter: Bruce Schuchardt
> Priority: Major
> Fix For: 1.10.0
>
> Time Spent: 1.5h
> Remaining Estimate: 0h
>
> In a system configured to use SSL in the cluster I observed many "Attempting
> TCP/IP reconnect" messages and failures to return "direct replies".
> I modified the ClusterCommunicationsDUnitTest to perform multiple updates in
> a thread and found the same behavior. The DistributionStats
> reconnectAttempts statistic showed new connections being formed when old ones
> were deemed corrupt. The retry logic in our communications caused the test
> to pass.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
