Open mikeharder opened 2 years ago
Could you use the container host or OS tools on the agent VM to put an upper memory/cpu bound on test-proxy?
Could you use the container host or OS tools on the agent VM to put an upper memory/cpu bound on test-proxy?
Outside of hypervisors, we haven't had much luck using software-based resource constraints (like container limits) for perf testing. The results are inconsistent and do not match the results from similar environments constrained by hardware or the hypervisor. But we could give it another try with the Azure SDK perf tests.
Add parameter to perf pipelines to use test-proxy instead of live resources. Compare running test-proxy on localhost with deploying to a service like Azure Container Instances (may require vnet peering with perf agent pool).