how to remove prometheus from Inbound in k8s

[RiverFerry]

i have installed linkerd-viz in my k8s cluster, i can view grafana dashboard by viz web, and latency here works great for me， it has defalut config:

histogram_quantile(0.5, sum(irate(response_latency_ms_bucket{namespace="$namespace", deployment="$deployment", direction="inbound"}[30s])) by (le, deployment))

my problem is, in my deployment A, inbound has multiple sources，like:

• B
• C
• D
• deploy/prometheus

deploy/prometheus exist in any of deployment, but i wish p50 latency can remove deploy/prometheus from inbound, so i can measure my own deployment app. is there has any solution? thanks!

[mateiidavid]

You can try to drop metrics with specific labels by modifying the prometheus scrape config.

linkerd2/viz/charts/linkerd-viz/templates/prometheus.yaml

Line 88 in 80b2fdb

- job_name: 'linkerd-proxy'

For example, the proxy job in the snippet above will configure prometheus to scrape the proxies' admin port. You might be able to drop labels based on regex using labeldrop?

[RiverFerry]

thanks for your reply, i will try this.