Monitors

kpulse ships 12 monitors, all enabled by default. Each is a key under monitors: in the ConfigMap. Set enabled: false to turn one off.

Monitor	Triggers	Severity	Page
pod_crashes	CrashLoopBackOff, OOMKilled, ImagePullBackOff, etc.	critical	Pod Crashes & Events
pod_restarts	> 5 restarts in 15 min	warning	Pod Crashes & Events
warning_events	Warning-type k8s Events	info	Pod Crashes & Events
pvc_usage	PVC > 80% warn, > 90% crit	warn/crit	PVC Monitoring
node_conditions	DiskPressure, MemoryPressure, PIDPressure, NotReady	critical	Node Monitoring
node_disk	rootfs/imagefs > 85% warn, > 92% crit	warn/crit	Node Monitoring
tls_cert_expiry	< 14 d warn, < 3 d crit	warn/crit	TLS Cert Expiry
rollout_stuck	Deployment/StatefulSet stuck > 15 min	warning	Workload Health
job_failed	Job condition Failed=True	warning	Workload Health
cronjob_missed	> 2 missed schedules	warning	Workload Health
hpa_at_max	HPA at maxReplicas > 30 min	warning	Workload Health
daemonset_unscheduled	desired != ready > 10 min	warning	Workload Health

Tuning rule of thumb

The defaults are calibrated to be quiet on a healthy cluster.

• Getting spammed? Raise the threshold, or raise dedupe.window, or add the noisy reason to warning_events.reasons_ignore.
• Something broke and kpulse stayed silent? Lower the threshold. Check if the namespace is in namespaces.exclude.
• Want everything in one daily summary? Set dedupe.digest.interval: 24h and add critical to dedupe.digest.severities (default excludes it so critical bypasses the digest).

Disabling a monitor

monitors:
  warning_events:
    enabled: false

Then kubectl -n kpulse rollout restart deploy/kpulse.