stress/

Route Controller Stress Tests

Proposed Metrics

• Create 1000 routes, then deploy RouteController and measure how long it takes for all the VirtualServices and Services to be created?
• Once there are 1000 routes (and their corresponding VirtualServices and Services), if we add 100 more routes quickly, how long does it take to create the VirtualService and Service?
• Given I have 1100 routes, how long does it take to modify 100 of routes as individual updates?
• Given I have 1100 routes, how long does it take to modify 1000 in bulk?
• Given I have 1100 routes, how long does it take to remove 100 routes ?
• Given I have 1000 routes, how long does it take to remove all the routes?

For each of these 6 metrics, we measure the time that each change takes to fully propagate by observing modifications to VirtualServices and Services, then the avg propagation of 95% of the new routes is under a baseline. If we fall outside of some tolerance range of the baseline, the test fails.

Architectural Details

• runs on KIND
• only deploy the Route CRD and the VirtualService CRD from Istio
• No Istio
• No deployment of cf-for-k8s
• No AIs
• deploys RouteController as part of each batch of tests

Run the tests

cd cf-k8s-networking/routecontroller/stress
ginkgo .