[teiid-issues] [JBoss JIRA] (TEIID-2172) Load balancing configuration can result in premature session termination
[
https://issues.jboss.org/browse/TEIID-2172?page=com.atlassian.jira.plugin...
]
Steven Hawkins resolved TEIID-2172.
-----------------------------------
Resolution: Done
opted for proactively disassociating of the session via assertIdentity(null). that does
add overhead to load-balancing as an additional synch remote call is being made. on the
other hand it means that server will rarely accumulate invalid sessions.
another approach would have been to not do a logoff in the socketclientinstance
disconnection, however that would lead more orphan sessions waiting to be cleaned up via
the sessionservice monitor.
this behavior could be made configurable later if seems necessary to remove the extra
remote call.
also tightened ping monitoring bounds on the server side to kill after 3 missed pings
instead of 5 and to monitor every 4 minutes instead of 10.
Load balancing configuration can result in premature session
termination
------------------------------------------------------------------------
Key: TEIID-2172
URL: https://issues.jboss.org/browse/TEIID-2172
Project: Teiid
Issue Type: Bug
Components: JDBC Driver, Server
Affects Versions: 7.4.4
Environment: Clustered environment with load balancing enabled for the Teiid data
source
Reporter: Marc Shirley
Assignee: Steven Hawkins
Priority: Blocker
Fix For: 8.2
When set up in a load balancing configuration with "SET NEW INSTANCE" as the
check valid connection SQL in the datasource definition, repeated messages indicating an
invalid session are created, similar to:
2012-08-24 10:39:58,818 WARN [org.teiid.SECURITY] (New I/O server worker #1-3)
Processing exception 'The specified session ID "5Wt/6zzmsLy2" is invalid. It
cannot be found in the userbase.' for session R8ZxYqjZzf/S. Exception type
org.teiid.client.security.InvalidSessionException thrown from
org.teiid.services.SessionServiceImpl.getSessionInfo(SessionServiceImpl.java:315). Enable
more detailed logging to see the entire stacktrace.
Not sure if this would be a defect in regards to the conditions that are resulting in the
messages to be generated (ie, the old session id shouldn't be being reused), or an
enhancement to clarify or adjust the log message or priority to indicate whether this will
actually affect the query being submitted.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira