- This app is created to reproduce mis-balancing in HAproxytech's ingress controller. It only servs
/serverIdendpoint which returns a json with serverId key and a random string which is the unique identifier for that server. For example:{serverId: ABCDEF} - Ingress object of demoapp is configured like following:
ingress: enabled: true className: "haproxy" annotations: haproxy.org/timeout-tunnel: "3600s" haproxy.org/backend-config-snippet: | balance url_param RouteToken check_post hosts: - host: test.demoapp paths: - path: /serverId pathType: Prefix
- HAproxy is configured to balance based on
url_param RouteToken
-
Install HAproxy Ingress Controller with Nodeport service type
-
Install demoapp using chart given in demoapp repository
helm install --create-namespace --namespace demoapp demoapp path/to/chart/given/in/repo
- It is simple script that creates a hashmap serverId<->routeToken map. Here routeToken also is random string which signifies to which pod the request should endup. The hope is whenever there is
RouteTokenparameter in url HAproxy should always send the request to appropriate pod. - For example, If script creates serverid map like following:
serverId<->routeToken ABC<->123 PQR<->456 XYZ<->789
- Then request with url
http://test.demoapp:30080/serverId?RouteToken=123should always endup on demoapp pod withserverId=ABC. The problem it deosn't happen number when pods are very high for example 50 demoapp pods
- Then request with url
- There is function in
test-ha.pytest_serverid_mismatch which tests the above explain behaviour - To run the script
python test-ha.py http://test.demoapp:<whatever-nodeport-you-have-configured>
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent