Replies: 3 comments 3 replies
-
@notcool11, thanks for your question. The host information is gathered from the available resource attributes for the batch/trace/metric data: . As long as both metric and trace pipelines utilize the same resource/resourcedetection processors that result in the matching relevant resource attributes the correlation should apply. The Splunk distribution of the Collector has a writeup about this when using your own configuration and not the defaults that we provide: https://github.com/signalfx/splunk-otel-collector/blob/main/docs/apm-infra-correlation.md.If you're still not seeing the desired effect would you mind providing a redacted config in a new issue in our project? You could also reach out to Splunk support at any time. |
Beta Was this translation helpful? Give feedback.
-
Right on, thanks for the quick reply!! Just updated our collector version and will set the yaml as the docs are recommending. Will post back with results. |
Beta Was this translation helpful? Give feedback.
-
Okay so a little good and a little confused. Made the update to add Now we can see Host metrics in the "Related Content" tabs on the bottom of Splunk APM and also Host Metrics on the default Service dashboard. Sweet!! Now we want more!! :) What does it take to get other types of related content to show up?? The default correlation is on Thanks!! |
Beta Was this translation helpful? Give feedback.
-
Looking at the sfx exporter, specifically the AddSpans function
This looks like the place where trace and metric data are correlated together. Need some guidance and/or told what I'm doing is stupid and will never work.
We have a Trace Collector (Splunk version 0.31.1) deployed in OpenShift that exposes the otlp receiver. We can send traces to that thing and using the signalfx exporter can see the traces showing up in Splunk Observability. Sweet!!
We have a separate Metrics Collector (Splunk version 0.31.1) deployed in OpenShift as a daemonset that pulls the metrics and sends to Splunk Observability as well. Also sweet!!
Now we are trying to correlate the traces with the metrics and not having a lot of luck. From looking at the code it seems like the trace/metric correlation relies on the trace having
k8s.pod.id
andk8s.container.id
as the default correlation. But it also looks like it is comparing to aResource
list??In our case this will never match because the Trace and Metric collectors are separate implementations. How do we proceeded?? Do we have to have the traces flow through the same collector that gathers the metrics?? Would love to keep them separate but if that is not possible we'll need to reconsider our design. Thanks.
Beta Was this translation helpful? Give feedback.
All reactions