Open KKHYeon opened 7 months ago
@thomasvn, is this a Helm issue or something else?
@KKHYeon @chipzoller Thanks for pointing out the issue! This seems to be a bug in the cost-model
logic, as opposed to the Helm chart.
Thanks, Thomas. Transferred to the features-bugs repository.
This is one location in the code, that matches the incorrect labels you point out. However, git blame
shows that these lines haven't been modified in ~8months. It's possible that these labels changed when network-costs moved from a Go based image, to a Rust based image. cc @mbolt35
Thank you for your response!
This issue seems related to the move to rust. The rust prom metric writer is likely just leveraging the internal struct field names (which use the _ naming... go used the camel casing).
This is a great catch, and thankfully an easy fix. Thanks for flagging me down @thomasvn - this one is definitely on me!
Also, to be clear, this is what I would consider a major feature breaking bug. I'm going to address this and release 0.17.3 ASAP, likely followed by a helm chart update.
This should be resolved with: https://github.com/opencost/opencost/pull/2478
Kubecost Helm Chart Version
1.108.1
Kubernetes Version
1.28
Kubernetes Platform
EKS
Description
In the Cost Analyzer pod, when querying related to network costs, it uses the ‘sameRegion/sameZone’ label for the query. However, the actual metrics collected are set and collected as ‘same_region/same_zone’. This could potentially lead to discrepancies or issues due to the difference in label naming conventions used in querying and data collection.
Steps to reproduce
Expected behavior
Cost Analyzer pod should use the same labels.
Impact
No response
Screenshots
No response
Logs
No response
Slack discussion
No response
Troubleshooting