-
Bug
-
Resolution: Unresolved
-
Major
-
None
-
4.16.z
-
Important
-
Yes
-
Rejected
-
False
-
Description of problem:
Openshift 4.16 Router (HAProxy 2.8) is opening new connections to dead pods after reload
Version-Release number of selected component (if applicable):
OCP 4.16.z and HAProxy 2.8
How reproducible:
Only on customer cluster
Steps to Reproduce:
1. Perform API Calls to the version X of the application
2. Start a rollout of the new version X+1 with `oc rollout latest myapp`
3. Perform API Calls to the version X+1 of the application, one or more API Calls may return 503 error
Actual results:
Router pod has old thread pointing to the configuration before the reload, a keep-alive connection may still communicate with old dead pods
Expected results:
Router pod needs to terminate keep-alive connections to dead pod
Additional info:
The issue may be related to the https://issues.redhat.com/browse/OCPBUGS-32044 , however customer is using OCP 4.14.34 and 4.14.37 without experiencing this issue
