opentelemetry-collector
opentelemetry-collector copied to clipboard
[request help]some questions about collector performance test
trafficstars
I read the doc about performance from this link, and i tried to reproduce the result. So i start a collector in k8s with 1c 2gi, and start the load test.
I used a simple config:
receivers:
otlp:
protocols:
grpc:
http:
exporters:
logging:
loglevel: info
extensions:
health_check:
pprof:
zpages:
service:
extensions: [health_check,pprof,zpages]
pipelines:
traces:
receivers: [otlp]
exporters: [logging]
metrics:
receivers: [otlp]
exporters: [logging]
logs:
receivers: [otlp]
exporters: [logging]
and set the submission rate to 100000 spans per second.
When the test starts, i tried to observe the metrics from localhost:8888.I found the number of the otelcol_receiver_accepted_spans grows too fast. (About 50000 spans per second , larger than 12k) And when the usage of the cpu and memory is also strange:
[root@vm5 load-test]# kubectl top pod -n otel otel-collector-7b6f65ffcd-jhbk9
NAME CPU(cores) MEMORY(bytes)
otel-collector-7b6f65ffcd-jhbk9 112m 54Mi
The uasge seems too low!
Did i make some mistake? Or the load test is not starting in a correct way?