CPU metrics

[amandahla]

Description

While monitoring Synapse, I noticed CPU usage has values different from those expected.

Shouldn't process_cpu_seconds_total be the sum of system and user usage?

values that are different from those

Steps to reproduce

• Install Synapse
• Collect metrics using Prometheus
• Verify last values for process_cpu_seconds_total, process_cpu_system_seconds_total and process_cpu_user_seconds_total

Homeserver

another

Synapse Version

1.112.0

Installation Method

Other (please mention below)

Database

PostgreSQL

Workers

Multiple workers

Platform

Ubuntu/Kubernetes

Configuration

No

Relevant log output

Not relevant

Anything else that would be useful to know?

Multiple core environment

[amandahla]

I'm closing this issue after finding out that the reason was another target writing the same metric with the same labels, causing a weird behavior for the rate function.