Skip to content

Diagnose

Diagnose is KubeManta's triage surface. The Diagnose tab itself is Free — its rule-based views (Failure Cascade, resource relationships, Health) need no license and no LLM key. Only the AI Diagnosis sub-tab is Pro.

What it does

Free (rule-based, no AI):

  • Failure Cascade Visualizer — a leveled dependency graph (root cause → blocked → indirectly affected) for outages that span multiple resources
  • Resource relationships — owner chain (Pod → ReplicaSet → Deployment), which Services select a workload, and PVC status/binds
  • Health — rule-based health signals per namespace/resource

Pro (ai, needs your LLM key):

  • AI Diagnosis (namespace scan) — sweep a namespace for unhealthy pods and get an AI-prioritized, explained summary
  • AI pod analysis — a focused AI root-cause explanation for a single failing pod (logs, events, resource requests/limits)
  • Results are cached per resource with a timestamp, so re-opening Diagnose doesn't always trigger a fresh LLM call

The AI sub-tabs respect the platform's AI Guardrails — data residency, cost caps, and the master AI on/off switch all apply. When unlicensed (or AI is off), the AI sub-tab shows a Pro upsell card while the rule-based views keep working.

Scheduled AI SRE reports (Pro, ai): the Reports sub-tab configures recurring, AI-generated cluster health reports — pick a schedule and namespaces (a filterable multi-select of discovered namespaces), and a background scheduler gathers the same read-only namespace state the AI diagnosis path uses and runs one guardrail-enforced summarization pass. Schedules are real wall-clock schedules — Daily / Weekly on a weekday / Monthly on a day, at HH:MM, plus cron and the legacy every-N-hours mode, all timezone-aware — and a restart never double-fires a slot. Reports are stored (age-pruned), viewable in-app, exportable as a formatted PDF (real headings, lists, tables, code blocks — not raw markdown), and can be delivered through any channel enabled under Alerts → Integrations (Slack, SMTP, PagerDuty, OpsGenie, Datadog, CloudWatch, webhook) — one failing channel never blocks the others. Every report rides the same guardrails as any other AI call (data residency, cost caps, prompt audit, redaction).

How to get there

Click Diagnose in the sidebar (AI group), or click Diagnose directly from an unhealthy pod row in Workloads.

Key actions

  • Use the cascade view (Free) when one failure appears to be causing several others
  • Check resource relationships (Free) to see a workload's owner chain, Services, and PVCs
  • Run an AI namespace scan (Pro) to triage everything at once
  • Drill into a single pod for AI pod analysis (Pro)
  • AI Diagnosis is disabled for pods that are already Running/Ready with zero restarts, or Succeeded/Completed — there's nothing to diagnose

Screenshots coming

A walkthrough of the namespace scan, Pod Failure Visualizer, and Failure Cascade view will be added here in a follow-up pass.