Diagnostics
Know the numbers are complete before you share them
Every cluster gets an hourly heartbeat and 10 checks, from the agent reporting to the cost being allocated. When something is missing, KubeOn names the cause and the fix.
Per-cluster checks
Ten checks on every cluster, every five minutes
Each cluster page shows the last 48 hours of snapshots, the agent version, run time and peak memory, and the result of 10 checks: agent reporting, snapshot ingestion, Kubernetes API inventory, resource quantities, metrics-server usage, Prometheus history, volume usage, billing data, nodes matched to billing and cost allocation.
- Agent reporting warns after 2 hours, errors after 6
- Nodes matched to billing is OK at 90% or more
- Latest snapshot contents, row by row
Coverage
Clusters you pay for but cannot see
KubeOn compares the clusters in your bill with the clusters that send snapshots. Any cluster that appears in the bill but has never reported is flagged as billed, no collector, on the Clusters page and in Diagnostics, so coverage gaps show up before someone asks why a team's numbers look low.
- Active, inactive and unmonitored clusters
- Environment detected from the cluster name
- A 30-day split cost backfill when a new cluster starts reporting
Data health
Every pipeline blocker named, with its fix
Data health follows each stage from agent to storage, inventory, hourly samples, billing and allocation. Blockers have names, such as a billing integration that failed, snapshots that stopped or provider IDs that do not match the bill, and each one comes with the remedy.
- Refreshed every 2 minutes
- Node-to-billing match rate
- Ingestion run history
Where the Kubernetes savings are
off the monthly bill: $21,842 a month found at our sample company
Sample company, not a customer result
Repair actions
Fix data problems without a support ticket
| Action | What it does |
|---|---|
| Snapshot backfill and retry | Re-ingests deferred or failed snapshots while they are in storage, which keeps them 14 days |
| Re-run allocation | Recomputes allocation for a window of up to 92 days. Days in the last 7 are also repaired automatically when late data arrives |
| Billing backfill | Re-reads 1 to 366 days of billing from the export |
| Integration reset | Clears a failed billing integration so it can reconnect |
Repair actions are limited to admins and platform engineers, and every run is logged.
Operational confidence for a cost platform
48-hour heartbeat
One cell per hour: ingested, partial, failed, deferred or missing.
Ten checks
From agent reporting to allocation, per cluster.
Billed, no agent
Clusters in the bill that never reported.
Agent telemetry
Version, run time and peak memory for every run.
Match rate
How much of each cluster joins to the bill.
Repair actions
Backfills and re-runs from the UI.
Self-healing billing
A failed billing integration retries every hour.
Platform checks
Agents, billing, encryption, email delivery and RBAC.
Diagnostics: your questions answered
KubeOn
Keep exploring
Evaluate diagnostics with your team
Bring your platform and FinOps priorities. We will walk through the data, controls and evidence for your evaluation.