Redesign EPP Metrics Pipeline to be Model Server Agnostic#461

Merged

k8s-ci-robot merged 19 commits intokubernetes-sigs:mainfrom

BenjaminBraunDev:metrics_refactor

Mar 14, 2025

Commits on Mar 13, 2025

start adding metrics changes for trion support
BenjaminBraunDev
committed
Refactor metrics to work with any prometheus metric naming convention based on EPP runtime flags.
BenjaminBraunDev
committed
Finalize metric refactor and testing.
BenjaminBraunDev
committed
Set streaming env var to false in triton ext_proc.yaml
BenjaminBraunDev
committed
Update titon server deployment to pull frozen repo branch instead of main for consistency.
BenjaminBraunDev
committed
Remove model server specific metric files and tests and point EPP image to main AR instead of testing registry.
BenjaminBraunDev
committed
Remove commented prints and old comments.
BenjaminBraunDev
committed
Remove triton support for now, make metrics mapping 1-to-1 with load balancing metrics.
BenjaminBraunDev
committed
moved files for cleaner diff
BenjaminBraunDev
committed
re-add todos and rename kv flag to reflect percentage usage.
BenjaminBraunDev
committed
Fix nits, move logging channel for backend/metrics.go from default to trace, fix comments.
BenjaminBraunDev
committed
Rebase into metric agnostic redesign.
BenjaminBraunDev
committed