feat(prometheus): enable container CPU throttling metrics collection

- Override default cAdvisorMetricRelabelings
- Remove cfs_throttled_seconds_total from drop regex
- Enables CPU Throttled panels in Grafana dashboards

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
2026-01-10 03:55:36 +09:00
parent 823edfbd88
commit c34f56945a

View File

@@ -25,6 +25,28 @@ kubelet:
replacement: "mayne-cluster"
- sourceLabels: [__metrics_path__]
targetLabel: metrics_path
# Override default drops to keep throttling metrics
cAdvisorMetricRelabelings:
# Drop unnecessary CPU metrics (but keep cfs_throttled_seconds_total)
- action: drop
regex: "container_cpu_(load_average_10s|system_seconds_total|user_seconds_total)"
sourceLabels: [__name__]
# Keep other default drops
- action: drop
regex: "container_fs_(io_current|io_time_seconds_total|io_time_weighted_seconds_total|reads_merged_total|sector_reads_total|sector_writes_total|writes_merged_total)"
sourceLabels: [__name__]
- action: drop
regex: "container_memory_(mapped_file|swap)"
sourceLabels: [__name__]
- action: drop
regex: "container_(file_descriptors|tasks_state|threads_max)"
sourceLabels: [__name__]
- action: drop
regex: "container_spec.*"
sourceLabels: [__name__]
- action: drop
regex: ".+;"
sourceLabels: [id, pod]
# Resource metrics
resourceRelabelings:
- targetLabel: cluster