Open dloucasfx opened 5 months ago
Pinging code owners:
See Adding Labels via Comments if you do not have permissions to add labels yourself.
This PR "does not" fix the main ask here, but improves and avoid the issue when the metric is not enabled
Removing needs triage
based on original PR being merged, I'll defer to code owners for more discussion on the potentially adding configuration options here.
Is this only an issue for windows or also Linux?
This issue has been inactive for 60 days. It will be closed in 60 days if there is no activity. To ping code owners by adding a component label, see Adding Labels via Comments, or if you are unsure of which component this issue relates to, please ping @open-telemetry/collector-contrib-triagers
. If this issue is still relevant, please ping the code owners or leave a comment explaining why it is still relevant. Otherwise, please close it.
Pinging code owners:
See Adding Labels via Comments if you do not have permissions to add labels yourself.
Is this only an issue for windows or also Linux?
Since the bug is in a WMI call, it looks like it's Windows only.
Another user hitting this issue, can we look into this?
Component(s)
receiver/hostmetrics
What happened?
Description
The
cpu.Counts
gopsutil func, which is called by the cpu scraper, does not set a deadline/timeout on its context, which forces WMIQueryWithContext to set it using the hardcoded timeout value of 3 seconds. In large busy env or/and low resourced, the wmi call can take longer than 3 seconds, which will lead to a context deadline exceeded error and fail to get the CPU counts.Steps to Reproduce
Find a windows host where the wmi calls take longer than 3 seconds and run the hostmetrics receiver with the cpu scraper.
Expected Result
Get all the metrics, including the physical and logical CPU counts
Actual Result
CPU counts are missing and we see this error in the logs
Collector version
v0.95.0
Environment information
Environment
OpenTelemetry Collector configuration
No response
Log output
Additional context
Suggestion is to use
CountsWithContext
instead ofCounts
and introduce a wmi_timeout option for cpuscrapercc: @atoulme who helped with the RCA