[
https://issues.apache.org/jira/browse/GEODE-8020?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17096875#comment-17096875
]
ASF GitHub Bot commented on GEODE-8020:
---------------------------------------
bschuchardt commented on pull request #5027:
URL: https://github.com/apache/geode/pull/5027#issuecomment-622034391
@pivotal-jbarrett yes, the performance issue would be back, **but** the
cluster would not hang. That is currently the case with 1.10, 1.11 and 1.12.
They will hang. It seems to only occur during HA events. The performance drop
is relative to builds containing the change that caused this hang. There are
no CI tests demonstrating the hang, so we released several times with this bug.
I can't leave in the functional fix w/o failing benchmark tests. Taking out
the functional fix lets the pipeline pass while I work on a new fix that
doesn't degrade performance.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
> buffer corruption in SSL communications
> ---------------------------------------
>
> Key: GEODE-8020
> URL: https://issues.apache.org/jira/browse/GEODE-8020
> Project: Geode
> Issue Type: Bug
> Components: membership, messaging
> Reporter: Bruce J Schuchardt
> Assignee: Bruce J Schuchardt
> Priority: Major
> Fix For: 1.12.1, 1.13.0
>
>
> When running an application with SSL enabled I ran into a hang with a lost
> message. The sender had a 15 second ack-wait warning pointing to another
> server in the cluster. That server had this in its log file at the time the
> message would have been processed:
> {noformat}
> [info 2020/04/21 11:22:39.437 PDT <P2P message reader for
> rs-bschuchardt-1053-hydra-client-1(bridgegemfire4_host1_12599:12599)<ec><v1>:41003
> unshared ordered uid=354 dom #2 port=55262> tid=0xad] P2P message
> reader@2580db5f io exception for
> rs-bschuchardt-1053-hydra-client-1(bridgegemfire4_host1_12599:12599)<ec><v1>:41003@354(GEODE
> 1.10.0)
> javax.net.ssl.SSLException: bad record MAC
> at sun.security.ssl.Alerts.getSSLException(Alerts.java:214)
> at sun.security.ssl.SSLEngineImpl.fatal(SSLEngineImpl.java:1728)
> at sun.security.ssl.SSLEngineImpl.readRecord(SSLEngineImpl.java:986)
> at sun.security.ssl.SSLEngineImpl.readNetRecord(SSLEngineImpl.java:912)
> at sun.security.ssl.SSLEngineImpl.unwrap(SSLEngineImpl.java:782)
> at javax.net.ssl.SSLEngine.unwrap(SSLEngine.java:626)
> at
> org.apache.geode.internal.net.NioSslEngine.unwrap(NioSslEngine.java:275)
> at
> org.apache.geode.internal.tcp.Connection.processInputBuffer(Connection.java:2894)
> at
> org.apache.geode.internal.tcp.Connection.readMessages(Connection.java:1745)
> at org.apache.geode.internal.tcp.Connection.run(Connection.java:1577)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> Caused by: javax.crypto.BadPaddingException: bad record MAC
> at sun.security.ssl.InputRecord.decrypt(InputRecord.java:219)
> at
> sun.security.ssl.EngineInputRecord.decrypt(EngineInputRecord.java:177)
> at sun.security.ssl.SSLEngineImpl.readRecord(SSLEngineImpl.java:979)
> ... 10 more
> {noformat}
> I bisected to see when this problem was introduced and found it was this
> commit:
> {noformat}
> commit 418d929e3e03185cd6330c828c9b9ed395a76d4b
> Author: Mario Ivanac <[email protected]>
> Date: Fri Nov 1 20:28:57 2019 +0100
> GEODE-6661: Fixed use of Direct and Non-Direct buffers (#4267)
> - Fixed use of Direct and Non-Direct buffers
> {noformat}
> That commit modified the NioSSLEngine to use a "direct" byte buffer instead
> of a heap byte buffer. If I revert that one part of the PR the test works
> okay.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
