My point exactly. How is the reply getting back to the TCP Client Receive node?
Original Message:
Sent: Thu May 09, 2024 05:56 PM
From: Philip Muchemi
Subject: TCP/IP Stress Test On App connect CP4I
@Francois Brandelik The load balancer is a global load balancer on openshift. So rather than using IP address of the worker nodes where the application is running I'm just using the Load balancer IP address.
------------------------------
Philip [Muchemi] [Integration Consultant]
[Integration Specialst]
[Suluhisho Systems Limited]
[Nairobi] [Kenya]
Original Message:
Sent: Thu May 09, 2024 05:52 PM
From: Francois Brandelik
Subject: TCP/IP Stress Test On App connect CP4I
So the input nodes are load balanced... how about the TCP Client Receive node?
------------------------------
Francois Brandelik
Original Message:
Sent: Thu May 09, 2024 02:33 PM
From: Philip Muchemi
Subject: TCP/IP Stress Test On App connect CP4I
Dear François,
Thank you for your prompt response. It seems there may have been a misunderstanding. Regarding CP4I, there is only one instance running, akin to the setup on the local server.
The message flow comprises three TCP Server input nodes, each utilizing different ports to accept requests. These nodes then send and receive messages to and from a backend application using TCP Client out and TCP Client receive nodes, respectively.
To enable client applications to send requests to the application via IP address and port, I have created three services of type NodePort, each corresponding to the ports at which the TCP Server input nodes are listening. Thus, to send a request to the application, one can utilize the load balancer IP along with any port that the TCP Server input nodes are employing.
Upon stress testing the application using JMeter on the local integration server, successful results were obtained. However, on the CP4I environment on OpenShift, half of the tests fail. I trust this clarification provides a clearer overview of the situation. Thank you.
------------------------------
Philip Muchemi
Original Message:
Sent: Wed May 08, 2024 04:50 AM
From: Francois Brandelik
Subject: TCP/IP Stress Test On App connect CP4I
Hi Philip,
If I understand you correctly in OpenShift you have 3 instances configured whereas you have only a single local integration server.
What happens if you decrease the number of pods in OpenShift to 1?
This is a flow where there is affinity and the TCP Client receive needs to be connected to the TCP Client out. With multiple instances (3) and a load balancer using the same TCP external route for all 3 instances, the relation between TCP Client receive and TCP Client out is broken and leads to 2/3rds of failures by design
Enjoy
------------------------------
Francois Brandelik
Original Message:
Sent: Tue May 07, 2024 02:34 PM
From: Philip Muchemi
Subject: TCP/IP Stress Test On App connect CP4I
Hi Team,
I am facing an issue where stress testing an application which utilizes the TCP/IP protocol on the IBM App Connect local integration server, yields successful results e.g for 100 requests. However, when the same tests are performed on an integration server deployed in the CP4I OpenShift environment, half of the tests are failing or rather most of them. The application is using the following nodes:
TCP Server input - Listens for incoming requests.
TCP Client out - Routes incoming requests to a backend application that accepts TCP/IP requests.
TCP Client recieve - Recieves the response from the backend application.
TCP Server out - Sends the response back to the client.
On Openshift I have configured three Node Port services for the TCP Server input nodes, since the application is listening on three TCP Server input nodes just for customization purposes. However when stress testing the application on cp4i environment the tests fail. How could I resolve this issue?
------------------------------
Philip Muchemi
------------------------------