PQ96684: CUSTOMER EXPERIENCES INCORRECT ROUTING OF REQUESTS TO NODEAGENTS, CAUSING AN OUT OF MEMORY CONDITION. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() APAR status Closed as program error. Error description Cust receives corba marshall error, unable to send cancel request when an application is formed betwen 2 application servers. More details below: - cust is using a Cisco machine, Windows OS - has 1 installation of WebSphere, 1 node as parent and 2 nodes as children - first node has client deployed called clusterGroup10a - 2nd node has EJB deployed called clusterGroup10b - client call a business method though RMI, when do this, their code does looping which calls 100 times which should result 100 results or data being requested. Apparently, when they get to the 50th call, it gives them this corba exception (mentioned in the beginning of the pmr). - the 2nd appServer which is in the 2nd node puts message on WAS MQ queue - memory used before the call was 90 mb - memory after the call was 320 mb (excessive?) - request was made - 12:35:11:472 - response was received - 12:50:28:521 Development's evaluation: Per ORB's evaluation that the request calls were being sent to the cluster server in CF4 but they were being sent to the nodeagent in CF6. This is a bug in WLM code that we push a wrong target onto the request stack, sent a "Forward invocation is not wlm enabled" message but did not pop this server off the stack so requests keep diverting to the nodeagent.Local fix Patch has been provided to customer, and an official iFix will be available as well which would be in 5.0.2.9.Problem summary **************************************************************** * USERS AFFECTED: WebSphere Application Server users of * * Workload Management (WLM) * **************************************************************** * PROBLEM DESCRIPTION: Customer experiences incorrect * * routing of requests to nodeagents, * * causing an out of memory condition * **************************************************************** * RECOMMENDATION: * **************************************************************** Per ORBs evaluation, the request calls were being sent to the cluster server in CF4 but they were being sent to the nodeagent in CF6. This is a bug in WLM code that we push a wrong target onto the request stack, sent a "Forward invocation is not wlm enabled" message but did not pop this server off the stack so requests keep diverting to the nodeagent.Problem conclusion Fix code to properly pop nodeagent requests This APAR is currently built and is included in the 5.0.2.9 fixPak, and is targeted for inclusion in the 5.1.1.2 fixPak Please refer to the Recommended Updates page for delivery dates: http://www-1.ibm.com/support/docview.wss?rs=180&context=SSEQTP &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: 00W
Software edition:
Reference #: PQ96684
IBM Group: Software Group
Modified date: Dec 2, 2004
(C) Copyright IBM Corporation 2000, 2008. All Rights Reserved.