Closed siggy closed 6 years ago
The perf-baseline environment runs Linkerd under load, with results observable in Prometheus / Grafana.
This change introduces a small Go process, reporter, that periodically queries Prometheus and prints a report to stdout.
Signed-off-by: Andrew Seigner siggy@buoyant.io
Depends on #233.
Example output:
$ kubectl -n l5d-perf logs -f $(kubectl -n l5d-perf get po --selector=app=reporter -o jsonpath='{.items[*].metadata.name}') time="2018-08-17T00:39:22Z" level=info msg="Waiting 7 minutes to publish first report for namespace l5d-perf..." Stats report: Protocol: h1: baseline-h1 Success rate: 100.00% Request rate: 187 p50 latency (us): 1134 p75 latency (us): 1404 p90 latency (us): 1740 p95 latency (us): 2054 p99 latency (us): 3148 p99 latency (us): 5435 Memory (bytes): 0 CPU (cores): 0.000 linkerd2 Success rate: 100.00% Request rate: 155 p50 latency (us): 1270 p75 latency (us): 1675 p90 latency (us): 2192 p95 latency (us): 2470 p99 latency (us): 3896 p99 latency (us): 5705 Memory (bytes): 0 CPU (cores): 0.000 Protocol: h2: baseline-h2 Success rate: 100.00% Request rate: 178 p50 latency (us): 384 p75 latency (us): 691 p90 latency (us): 1134 p95 latency (us): 1590 p99 latency (us): 3284 p99 latency (us): 7368 Memory (bytes): 0 CPU (cores): 0.000 linkerd2 Success rate: 100.00% Request rate: 191 p50 latency (us): 1500 p75 latency (us): 1781 p90 latency (us): 2338 p95 latency (us): 2802 p99 latency (us): 3861 p99 latency (us): 7583 Memory (bytes): 0 CPU (cores): 0.000
The perf-baseline environment runs Linkerd under load, with results observable in Prometheus / Grafana.
This change introduces a small Go process, reporter, that periodically queries Prometheus and prints a report to stdout.
Signed-off-by: Andrew Seigner siggy@buoyant.io
Depends on #233.
Example output: