Files
Myasnikov Daniil b758de9ace fix(dashboard): apply Running-phase join to per-namespace memory panels
The per-namespace Memory request/limit panels summed
kube_pod_container_resource_{requests,limits} over all pods regardless of
phase, while the sibling CPU panels were updated to multiply by
* on (namespace, pod) group_left(phase) (kube_pod_status_phase{phase="Running"} == 1).
Apply the same join so the paired CPU/Memory rows count the same pod population.

Signed-off-by: Myasnikov Daniil <myasnikovdaniil2001@gmail.com>
2026-06-24 15:29:24 +05:00
..
2025-04-22 17:42:15 +02:00
2025-04-22 17:42:15 +02:00
2024-09-26 11:37:32 +02:00
2024-02-08 12:04:32 +01:00
2024-02-08 12:04:32 +01:00
2025-04-22 17:42:15 +02:00
2024-02-08 12:04:32 +01:00