PK26039: ADJUSTMENTS MADE TO NETWORK CLOCK TIME CAUSING DUPLICATE UUID EVENTS WHEN EXECUTING THE CODE ON FAST PROCESSORS. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() APAR status Closed as program error. Error description The following exceptions are received in the SystemOut.log over and over [1/3/06 14:00:20:451 EST] 19d7f6d CScopeRootImp E CSCP0007E: An unexpected error occurred in method completeActivity in cl ass CScopeRootImpl; the exception stack trace follows: com.ibm.ws.javax.activity.NoActivityException com.ibm.ws.activity.UserActivityImpl.completeWithStatus(UserActi vityImpl .java:2233) com.ibm.ws.cscope.CScopeRootImpl.completeActivity(CScopeRootImpl .java:14 5) com.ibm.ws.cscope.CScopeRootImpl.afterCompletion(CScopeRootImpl. java:358 ) com.ibm.ws.Transaction.JTA.RegisteredSyncs.distributeAfter(Regis teredSyn cs.java:195) com.ibm.ws.Transaction.JTA.TransactionImpl.distributeAfter(Trans actionIm pl.java:2452) On a network-scale, the different devices connected within the network can and often do have a different perspective on what the actual current time presently is. However, within distributed systems this can cause a problem and therefore there exists a protocol for networked UNIX computers to provide some consensus of time and permits changes to local clocks on those devices. This infrastructure is implemented in UNIX via the network time process (NTP) daemon. And, the occassion of time going backwards represented the events of some corrections for those local clocks where that clock sequence moved backwards into the past - not by much, but by enough. Thus, the problem develops because in constructing the type of unique UUID being used for such global unique references, the current timestamp is one of the primary constituents that forms this unique value. This establishes the time dependency. Furthermore, the rate at which the JVM may process the code that seeks the creation of new unique references was just fast enough under Server JIT optimisation for the problem of duplicate timestamps to become an issue. Thus, the risk of the problem of obtaining a duplicate timestamp is related to the average size of the correction interval negotiated by the NTP time correction daemon processes.Local fix N/AProblem summary **************************************************************** * USERS AFFECTED: All WebSphere Application Server Activity * * Service Users * **************************************************************** * PROBLEM DESCRIPTION: Adjustments made to network clock time * * allowed the generation of duplicate * * Unique User ID's. This caused some * * activity sessions to be created with * * the same identifiers, spawning a * * variety of problems related to the * * management of activity sessions. These * * are usually accompanied by error msg * * CSCP0007E, caused by the occurrence of * * an unexpected error in the activity * * session. * **************************************************************** * RECOMMENDATION: * **************************************************************** On a network-scale, the different devices connected within the network can and often do have a different perspective on what the actual current time presently is. However, within distributed systems this can cause a problem and therefore there exists a protocol for networked UNIX computers to provide some consensus of time and permits changes to local clocks on those devices. This infrastructure is implemented in UNIX via the network time process (NTP) daemon. And, the occassion of time going backwards represented the events of some corrections for those local clocks where that clock sequence moved backwards into the past - not by much, but by enough. Thus, the problem develops because in constructing the type of unique UUID being used for such global unique references, the current timestamp is one of the primary constituents that forms this unique value. This establishes the time dependency. Furthermore, the rate at which the JVM may process the code that seeks the creation of new unique references was just fast enough under Server JIT optimisation for the problem of duplicate timestamps to become an issue. Thus, the risk of the problem of obtaining a duplicate timestamp is related to the average size of the correction interval negotiated by the NTP time correction daemon processes.Problem conclusion The UUID class was modified to detect when a value is returned for a time that is in the past, compared to the previously returned valid time. If this happens, the class waits until the network clock returns a time value which is later than the last previously returned valid time. If the network clock returns a time from the past (when compared to a previously returned time), it will output a message to the system log, indicating that the class will wait until a new valid time is returned. The fix for this APAR is currently targeted for inclusion in fixpacks 5.1.1.12, 6.0.2.17 and 6.1.0.4 Please refer to the Recommended Updates page for delivery information: http://www.ibm.com/support/docview.wss?rs=180&uid=swg27004980Temporary fix Comments
APAR is sysrouted FROM one or more of the following: APAR is sysrouted TO one or more of the following: Modules/Macros
Publications Referenced
|
Product categories: Software > Application Servers >
Distributed Application & Web Servers > WebSphere Application
Server > General
Operating system(s):
Software version: 10W
Software edition:
Reference #: PK26039
IBM Group: Software Group
Modified date: Sep 13, 2006
(C) Copyright IBM Corporation 2000, 2008. All Rights Reserved.