8000 knative queue proxy metric corrupted by different service monitor · Issue #15646 · knative/serving · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
knative queue proxy metric corrupted by different service monitor #15646
Closed
@yoona3316

Description

@yoona3316

Ask your question here:

Knative queue proxy metric is overmeasured when there's two service monitor connected to same pod.

In detail

There's pod named A with port 80, 9091 exposed. Pod A has two containers, one for queue proxy, the other for app container.

Service monitor A pull metrics from port 80, collects data defined by app container.
Service monitor B pull metrics from port 9091, collects data defined by knative metric.
Both service monitor pass through queue proxy.

The problem is, for service monitor B, it collects metric including API call from service monitor A.
For example, metric named revision_app_request_count counts API call from service monitor A, then the value is overmeasured.

Is there any way to exclude other service monitor API call from knative metric so that i can get actual value, not overmeasured or misguided one?

Metadata

Metadata

Assignees

No one assigned

    Labels

    kind/questionFurther information is requested

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      0