HTTP Keep alive for HTTP Routing assertion is maintaining the connection pool longer than expected

Products

• CA API Gateway

Issue/Introduction

For the API Gateway there is a requirement to control the time an outbound idle connection within the pool is maintained.

When attempting to change the setting io.httpConnectionIdleTimeout from 5s to 3s, network traces show that the encrypted alert from the gateway is taking between 500 ms to 3 seconds to reap the connection. How do you get a more prescriptive time between the CWP setting and when the gateway will close down the connection? It appears connections 'only clean the connection pool lazily'.

 

Environment

All supported versions of the CA API Gateway

Cause

The checker thread for idle connections needs to lower its checker time.

Resolution

There are Cluster Wide properties that will make the gateway close idle connections more aggressively. 

1. Update Cluster Wide properties:

io.httpConnectionIdleTimeout - set to 3500ms

io.httpConnectionIdleInterval - set to 1s

2. Restart the gateway after making these changes