Docs for exporting telemetry in LangSmith + Example OTEL Collector configuration #852
Conversation
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎
|
|
|
||
| ## LangSmith Services | ||
| The following LangSmith services expose metrics at an endpoint, in the Prometheus metrics format. | ||
| - <b>Backend</b>: `http://<backend_service_name>.<namespace>.svc.cluster.local:1984/metrics` |
There was a problem hiding this comment.
host-backend as well?
There was a problem hiding this comment.
and might we want to grab nginx metrics? i believe nginx has a way to export metrics also
| ## Redis | ||
| If you are using the in-cluster Redis instance from the Helm chart, LangSmith can expose metrics for you if you upgrade the chart with the following values: | ||
| ```yaml | ||
| redis: |
There was a problem hiding this comment.
like i mentioned before, I think we should include the exporters in the separate monitoring chart? Ideally dont want to include another image that people are worried about inside our chart?
There was a problem hiding this comment.
alternative crazy thought, what if we run an adhoc queue job to export redis/pg metrics -> otel endpoint (push vs pull model)
There was a problem hiding this comment.
sure let's move to standalone exporter instances as part of the monitoring chart for now.
The queue job seems like quite the task, can assess if people are against the exporter images and decide from there
There was a problem hiding this comment.
yea fair lets do that
First part of self hosted observability v1
Matching docs PR: langchain-ai/helm#353