Hello Obs team!
I’ve currently implemented a subordinate for “Grafana Alloy” to test out the concept shipping logs to loki.
Grafana Alloy, is a open telemery collector, with extra ability for grafana cloud. It is most likely drop in compatible with the standard/native otel-collector.
The reason we are not using the K8 version from Canonical is that we don’t use K8, but have a 100% machine platform.
I want to make sure tje Grafana Alloy charm passes on juju topology just like the cos charms does, but not 100% sure what is the correct version to use.
Question 1: What is the correct specc for a juju topoligy. I think I have seen a few variants.
Version 1:
“model”,“model_uuid”,“application”,“unit”,“charm_name”
Version 2:
“juju_model”,“juju_model_uuid”,“juju_application”,“juju_unit”,“juju_charm_name”
Question 2: What is it that I should pass on to grafana loki as labels? The topology as-is, or, rename the labels?
Question 3: Is there any repo which has the vm versions of the otel collector charms? I could only see the k8 one you linked earlier. I would benefit from trying to mimic the behaviour/implementation.
Question 4: Would it be possible to add some example on how to use the “from cosl import loki_logger”? I’m thinking to use this for sending logs about upgrade events for charms to a central place to track upgrades we do across many models, separate from workload logging. I want to track certain events, like upgrades, for all of our charms to get a good track record of what happens in pur environments.
A note: I wanted to use the cosl library, and found an error using it with machine charms on ubuntu24.04: ModuleNotFoundError: No module named 'puccinialin' · Issue #149 · canonical/cos-lib · GitHub - the workaround is quite nasty, but works.
Hope to collaborate more as we move along here.