retina icon indicating copy to clipboard operation
retina copied to clipboard

Enrich flows with Service name and namespace

Open anubhabMajumdar opened this issue 1 year ago • 6 comments

Is your feature request related to a problem? Please describe.

Flows with source/destination IP mapped to K8s services are not enriched with the name as it does for Pods.

Describe the solution you'd like A clear and concise description of what you want to happen.

Describe alternatives you've considered A clear and concise description of any alternative solutions or features you've considered.

Additional context Add any other context or screenshots about the feature request here.

anubhabMajumdar avatar Oct 30 '24 19:10 anubhabMajumdar

@anubhabMajumdar, is this the reason why I only get "unknown app" on everything in Hubble UI? If that is the case then I will create a new issue for that but if not, when can we expect this to be fixed? Having anything to go on would be great 😅

roberthstrand avatar Jan 09 '25 10:01 roberthstrand

Hi @roberthstrand , it's this issue as well as https://github.com/microsoft/retina/issues/1252 .

anubhabMajumdar avatar Jan 21 '25 15:01 anubhabMajumdar

I just installed the hubble version of the helm chart, and I can see metrics, but all the hubble_flows_processed_total metrics have either the source or the destination populated, but never both. Is this the same issue?

FrederikNJS avatar Jan 21 '25 15:01 FrederikNJS

@anubhabMajumdar any update on this ?

VincentS avatar Apr 04 '25 10:04 VincentS

I just installed the hubble version of the helm chart, and I can see metrics, but all the hubble_flows_processed_total metrics have either the source or the destination populated, but never both. Is this the same issue?

This is because we set the following context for metrics - https://github.com/microsoft/retina/blob/main/deploy/hubble/manifests/controller/helm/retina/values.yaml#L210-L213 . This essentially makes sure metrics will have either source or destination and never both. This ensures the cardinality of metrics is limited, and retina works with limited memory even in large scale clusters. If you need both source/destination labels, you can change the context and re-deploy. Hope this helps!

anubhabMajumdar avatar Jul 01 '25 16:07 anubhabMajumdar

@anubhabMajumdar any update on this ?

Hi @VincentS , this is still on our roadmap. The change is pretty involved, because we need to implement our own service cache or modify upstream Cilium service cache to interface with Retina parser.

anubhabMajumdar avatar Jul 01 '25 17:07 anubhabMajumdar