Every platform pod runs at priority 0. During the 2026-11-09 disk eviction the tech-log production pods were ranked with Grafana and kube-state-metrics. The kubelet ranks evictions by usage over requests and then by priority, so without classes the database and the production app have no head start. Add two classes and attach nothing yet: platform-critical (1000000, may preempt) for platform-postgres, Gitea, Keycloak and tech-log, and platform-low (-1000, never preempts) for Grafana, pgAdmin, Loki and Tempo. Neither is the global default, so no pod changes silently. Risk: creating the classes changes no pod. Attaching them restarts the pods, so that is left for the k3s restart window. Verification: render-tenants-and-policies.sh (added later in this series) checks the values, the preemption policy and globalDefault.
26 lines
743 B
YAML
26 lines
743 B
YAML
apiVersion: scheduling.k8s.io/v1
|
|
kind: PriorityClass
|
|
metadata:
|
|
name: platform-critical
|
|
labels:
|
|
app.kubernetes.io/part-of: platform
|
|
value: 1000000
|
|
globalDefault: false
|
|
preemptionPolicy: PreemptLowerPriority
|
|
description: >-
|
|
Stateful platform services and production workloads that other services depend on
|
|
(platform-postgres, Gitea, Keycloak, tech-log). Evicted after everything else.
|
|
---
|
|
apiVersion: scheduling.k8s.io/v1
|
|
kind: PriorityClass
|
|
metadata:
|
|
name: platform-low
|
|
labels:
|
|
app.kubernetes.io/part-of: platform
|
|
value: -1000
|
|
globalDefault: false
|
|
preemptionPolicy: Never
|
|
description: >-
|
|
Tools whose loss does not stop a user-facing service (Grafana, pgAdmin, Loki, Tempo).
|
|
Evicted first and never preempts another pod.
|