-
Bug
-
Resolution: Done
-
Normal
-
None
-
4.11
-
None
-
Moderate
-
None
-
Sprint 229, Sprint 230, Sprint 231
-
3
-
Rejected
-
False
-
Description of problem:
When routes are created more than 80, SSL connections between OAuth Proxy container and HAProxy are disconnected with the following error messages
2022/12/15 21:37:01 server.go:3120: http: TLS handshake error from 10.128.18.27:47142: write tcp 10.128.10.57:8443->10.128.18.27:47142: write: connection reset by peer
With Model serving with 100 connections made OAuth proxy container failed so model serving pod failed too.
Version-Release number of selected component (if applicable):
How reproducible:
Always
Steps to Reproduce:
git clone https://github.com/Jooho/jhouse_openshift.git cd jhouse_openshift/OAuthProxy/https-reencrypt oc new-project oauth-proxy oc new-app -S php -n oauth-proxy oc new-app --template=cakephp-mysql-example -n oauth-proxy oc apply -f ./ oc replace -f ./svc-cakephp-mysql-example.yaml oc scale dc/cakephp-mysql-example --replicas=2 # Wait until all pods are running. export Token=$(oc sa new-token user-one) export URL=$(oc get route cakephp-mysql-example -ojsonpath='{.spec.host}') curl -o /dev/null -I -w "%{http_code}" --silent --location --fail --show-error --insecure https://${URL}/ -H "Authorization: Bearer ${Token}" # Start reproducing the error cat <<EOF> /tmp/cakephp.yaml apiVersion: route.openshift.io/v1 kind: Route metadata: labels: app: cakephp-mysql-example template: cakephp-mysql-example name: cakephp-mysql-example namespace: oauth-proxy spec: port: targetPort: oauth-https tls: insecureEdgeTerminationPolicy: Redirect termination: reencrypt to: kind: Service name: cakephp-mysql-example weight: 100 wildcardPolicy: None EOF for i in {1..100} ; do sed "7s/name:.*/name: cakephp-mysql-example-$i/g" /tmp/cakephp.yaml |oc apply -f - ; done # Check the error oc logs dc/cakephp-mysql-example -c oauth-proxy
Actual results:
Disconnected connections between OAuth Proxy and HAProxy
Expected results:
No errors happen
Additional info:
When I set replica of Router to 1, the issue was gone. However, if I increased it to 3, the issue was still around. So I don't think it is resource limitation issue.
- is related to
-
RHODS-6447 RHODS Dashboard oauth-proxy crashing with 1000 users
- Closed
-
OCPBUGS-33533 Incorrect Load Balancing Algorithm Applied Due to Mismatched Ports in spec.port.to and Alternate Backend
- Closed
- links to