[BUG] Metric "opensearch_os_cpu_percent" not reporting a valid value

[JakeHardy14]

What is the bug?
Having just upgraded to 3.2.0, I can see that previous Grafana dashboards reporting on the value of "opensearch_os_cpu_percent" are now returning a value of -1.0 for all nodes in the cluster which makes it impossible to observe the CPU usage.

# HELP opensearch_os_cpu_percent CPU usage in percent
# TYPE opensearch_os_cpu_percent gauge
opensearch_os_cpu_percent{cluster="Stage",node="Stage-opensearch-data-node-0",nodeid="Vl0KPIV2QGGvq8fEv0Grjg",} -1.0

How can one reproduce the bug?
Steps to reproduce the behavior:

1. Install a 3.2.0 Opensearch Cluster
2. Add the 3.2.0 opensearch-prometheus-exporter
3. Inspect the output on {node}:9200/_prometheus/metrics
4. Observe the value output for "opensearch_os_cpu_percent"

What is the expected behavior?
The value should be a float between 0 and 100

What is your host/environment?

• OS: RedHat 8.10

[YeonghyeonKO]

I ran into the exactly same issue and remained an issue here: opensearch-project/OpenSearch#19120

[YeonghyeonKO]

Hi @ritvibhatt ,
Afaik the metrics from OpenSearch 3.2.0 show this issue but I am not sure the root cause is in line with prometheus-exporter. Do you have any ideas about this issue by any chance?

[andrross]

Catch All Triage - 1 2

[JakeHardy14]

Hi @andrross, are you able to share the outcome of the triage please?

[lukas-vlcek]

@JakeHardy14 the goal of the triaging meeting is to check formal aspects of the ticket, ie. if the ticket provides all the required information, etc... The goal of the triaging meeting is not to evaluate the technical details or troubleshoot the real issue. In this case the "result" of the triaging meeting was that the "untriaged" label was removed which means the ticket is not missing any formal information and should be worked on.

Maybe you noticed that things are starting to move again with this plugin. I will have a look at the issue. Thanks for reporting.