altquark
Legendary Poster
OK - this one has me stumped, and I have scratched my head long enough over this nasty issue today. No doubt something is screaming in my face - but I cannot figure it out !
Brief description : The issue is that we are in the midst of creating additional application servers - and for some reason, some application servers are able to hang the client when a kernel dies.
Versions : We are on Xe SP15.1, HPUX 11.1, Oracle 8.1.7 - the client is NT4.0SP6
Long description : We map a business function to run on APPSVR - where APPSVR is an Enterprise Server (ie has a Database and OW Logic on it). APPSVR has its own Servermap Datasource.
We perform an instruction against the business function, and note the JDENet_K process that runs. We allow the function to complete to 100%. We kill the process on the server using "kill"
We then repeat the instruction. The client attempts to call the function - but fails and provides an error to the client "You must exit all the way out of this application". We exit back to OEXPLORER and re-enter the application. When we repeat the call, the application has spawned a new JDENet_K on the server - and everything is wonderful !
Then we map (in OCM) the function to run on a true 3-tier application server. The application server does NOT have its own Server Map - it uses the server map for the Enterprise Server.
When we call the function, everything is fine. When we kill the kernel and recall the function, the client logs the same "error connect" messages to the JDE.LOG - returns "99" as a function error, and then goes into a loop. The CPU hits 100% on the client and we have to End Task in Task Manager to shut down OEXPLORE.EXE. We have left the process running for >2 hours just in case it returns (it never does).
FINAL test was to create a "spoof" name for the Enterprise Server. By creating a new Application Server in the client HOSTS file locally - and mapping a function to this new name, we were able to get the client to hang in the same manner as above. In fact, the only way to get the client to work correctly was to have the Enterprise Server name.
This is pretty nasty for us, since we are trying to roll out a number of application servers for failover - and was wondering if anyone had seen their clients/citrix machines "hang" because a Kernel process had unexpectedly died ? I went through the knowledge garden, and checked as much as I could but to no avail.
By the way, I also tried SP17.1 on the client, just to see if it were a problem that JDE had fixed without any publication - but this also failed.
Help please !
Thanks
Jon Steel
Xe Upgrade Specialist
erpSOURCING LLC
ERP Sourcing
http://www.erpsourcing.com
[email protected]
Brief description : The issue is that we are in the midst of creating additional application servers - and for some reason, some application servers are able to hang the client when a kernel dies.
Versions : We are on Xe SP15.1, HPUX 11.1, Oracle 8.1.7 - the client is NT4.0SP6
Long description : We map a business function to run on APPSVR - where APPSVR is an Enterprise Server (ie has a Database and OW Logic on it). APPSVR has its own Servermap Datasource.
We perform an instruction against the business function, and note the JDENet_K process that runs. We allow the function to complete to 100%. We kill the process on the server using "kill"
We then repeat the instruction. The client attempts to call the function - but fails and provides an error to the client "You must exit all the way out of this application". We exit back to OEXPLORER and re-enter the application. When we repeat the call, the application has spawned a new JDENet_K on the server - and everything is wonderful !
Then we map (in OCM) the function to run on a true 3-tier application server. The application server does NOT have its own Server Map - it uses the server map for the Enterprise Server.
When we call the function, everything is fine. When we kill the kernel and recall the function, the client logs the same "error connect" messages to the JDE.LOG - returns "99" as a function error, and then goes into a loop. The CPU hits 100% on the client and we have to End Task in Task Manager to shut down OEXPLORE.EXE. We have left the process running for >2 hours just in case it returns (it never does).
FINAL test was to create a "spoof" name for the Enterprise Server. By creating a new Application Server in the client HOSTS file locally - and mapping a function to this new name, we were able to get the client to hang in the same manner as above. In fact, the only way to get the client to work correctly was to have the Enterprise Server name.
This is pretty nasty for us, since we are trying to roll out a number of application servers for failover - and was wondering if anyone had seen their clients/citrix machines "hang" because a Kernel process had unexpectedly died ? I went through the knowledge garden, and checked as much as I could but to no avail.
By the way, I also tried SP17.1 on the client, just to see if it were a problem that JDE had fixed without any publication - but this also failed.
Help please !
Thanks
Jon Steel
Xe Upgrade Specialist
erpSOURCING LLC
ERP Sourcing
http://www.erpsourcing.com
[email protected]
