describe JVM CPU metrics #3487
describe JVM CPU metrics #3487
Conversation
| instrument: gauge | ||
| unit: "1" | ||
|
|
||
| - id: metric.process.runtime.jvm.system.cpu.load_1m | ||
| type: metric | ||
| metric_name: process.runtime.jvm.system.cpu.load_1m | ||
| brief: "Average CPU load of the whole system for the last minute." | ||
| note: > | ||
| The value range is [0,n], where n is the number of CPU cores. |
There was a problem hiding this comment.
Hum, did I get it wrong or the link provided means something else? I couldn't find any mention of CPU cores there. It says
is the sum of the number of runnable entities queued to the available processors and the number of runnable entities running on the available processors averaged over a period of time
Also, the link says when load average is not available, it should return a negative number, which contradicts with the value range added here.
There was a problem hiding this comment.
I added the part about negative values, but I still think "CPU cores" is equivalent to the longer explanation in the link. Can you elaborate when/how it could be different?
|
There's also this PR that renames some of these metrics, maybe that also impacts things here? #3472 |
|
@zeitlinger heads up - I'm closing this PR since semantic conventions are being removed from this repo (opentelemetry-specification), we'll ask you to resubmit the PR in a new repo, please refer to #3474 (comment). |
Improve description of jvm cpu metrics - especially the value ranges.