additional test dimensions in lifecycle test environment #225
Description
The test environment introduced in BuoyantIO/emojivoto#41, intended to exercising route, telemetry, and service discovery lifecycles in Conduit, includes the following:
- conduit control + data plane
- 1 slow-cooker, 100qps
- 10 HTTP point-to-point servers
- 10 gRPC terminus servers, with container restart every 600 requests (~1 minute)
We should extend this environment to exercise additional dimensions, namely:
- Introduce dynamic scaling of lifecycle env BuoyantIO/emojivoto#45 M number of point-to-point servers
- Introduce dynamic scaling of lifecycle env BuoyantIO/emojivoto#45 N number of terminus servers
- X call-depth length via point-to-point servers
- Introduce dynamic scaling of lifecycle env BuoyantIO/emojivoto#45 Y discrete call graphs
- Delete entire pods in lifecycle environment BuoyantIO/emojivoto#44 full pod restart, on an interval, rather than container
- Redeploy bb-terminus deployment in lifecycle env BuoyantIO/emojivoto#49 Deployment rolling restart, on an interval
- Introduce baseline and TLS configs to lifecycle BuoyantIO/emojivoto#51 injected vs. baseline performance
- performance at 10qps, 100qps, 1,000qps, 10,000qps
- Introduce baseline and TLS configs to lifecycle BuoyantIO/emojivoto#51 with and without TLS (depends on Name ClusterRole objects to be namespace-specific linkerd2#1295)
The Prometheus Benchmark Grafana dashboard introduced in linkerd/linkerd2#984 should provide better performance analysis of Prometheus in Conduit.
Relates to BuoyantIO/emojivoto#42.
(this issue copied from BuoyantIO/emojivoto#43)