[mlbiscoc]

https://issues.apache.org/jira/browse/SOLR-18544

Partially a regression from SOLR-18407 but also partially a JDK bug that got exposed from the OTEL upgrade.

What I found was running the PlacementPluginIntegrationTest.testAttributeFetcherImpl with 21.0.6-tem fails but running with 21.0.9-tem passes.

This was because previously the metric was sourced from JFR. Then after the upgrade it started to source it from JMX which exposed from JDK bug. So this change just sources it from JFR again.

[epugh]

I love me a java class class called "Experimental" ;-). I wonder if we need the long comment... "Get CPU utlizatoin from JFR instead of JDG to work aroudn bugs"?

[dsmiley]

Just approving based on what you're saying. I no longer have the older JDK 21 to try it

[nick-boss-tech]

Quick Review:
The JFR selector behaves as intended in the pinned OpenTelemetry version: it preserves the existing experimental metrics and suppresses the duplicate JMX CPU source.
P3 readiness issue: the changelog entry is missing, and the current-head validator fails for that reason. A short SOLR-18544 bug-fix entry should cover the operator-visible metric recovery.
Precommit and Crave tests are green. The older-JDK fix remains author-reported; the existing placement assertion checks metric presence, which alone does not prove that JFR samples update its value.
AI-generated static review.