CRD reference

Two cluster-scoped custom resources, group mirops.mirops.io/v1. You author UpgradeAnalysis; the operator creates RemediationPlan. Everything works with plain kubectl.

UpgradeAnalysis

Minimal

apiVersion: mirops.mirops.io/v1
kind: UpgradeAnalysis
metadata:
  name: to-1-34
spec:
  targetVersion: "1.34"   # the only required field

Full spec

spec:
  targetVersion: "1.34"             # REQUIRED
  scoringProfile: production        # production (default) | non-production

  scope:
    mode: all                       # all (default) | application
    excludeNamespaces: [monitoring]

  ai:
    enabled: true                   # default false
    provider: anthropic             # anthropic | openai
    model: claude-sonnet-4-6        # optional
    maxTokens: 2048                 # 1–32768, default 2048
    credentialsSecret: mirops-ai    # Secret in the operator namespace
    remediation:
      enabled: true                 # default false
      maxRiskLevel: low             # low (default) | medium | high
      autoApprove: false            # default false

  resync:
    interval: "30m"                 # empty = run once

  source:
    type: file                      # file (default) | s3 | blob | pvc
    # file/pvc: path
    # s3:   bucket, region, key, credentialsSecret
    # blob: accountName, containerName, blobName, credentialsSecret

spec fields

Field	Values	Default
targetVersion	string, e.g. "1.34"	required
scoringProfile	production | non-production	production
scope.mode	all | application	all
scope.excludeNamespaces	[]string	—
ai.enabled	bool	false
ai.provider	anthropic | openai	anthropic
ai.model	string	provider default
ai.maxTokens	int 1–32768	2048
ai.credentialsSecret	string	—
ai.remediation.enabled	bool	false
ai.remediation.maxRiskLevel	low | medium | high	low
ai.remediation.autoApprove	bool	false
resync.interval	duration	— (once)
source.type	file | s3 | blob | pvc	file

status (read-only)

Field	Meaning
decision	SAFE / WARNING / CRITICAL / ERROR
totalScore	readiness gauge (0–100)
reason	primary reason for the verdict
aiScore, aiReasoning, aiModel, aiError	AI results (when enabled)
addonsChecked, incompatibleAddons	compatibility counts
reportState, reportError, reportLocation	report write outcome
lastAnalysisTime	timestamp of the last run

RemediationPlan

Created by the operator when ai.remediation.enabled is true. You set spec.approved: true to execute.

apiVersion: mirops.mirops.io/v1
kind: RemediationPlan
metadata:
  name: to-1-34-remediation
spec:
  upgradeAnalysisRef: to-1-34
  approved: false                   # ← set true to execute
  actions:
    - id: action-1
      type: restart-pod             # restart-pod | delete-pod | scale-deployment | cordon-node
      namespace: default
      name: crash-test-...
      reason: "..."
      risk: low                     # low | medium | high
      skip: false
      params: {}                    # type-specific, e.g. { replicas: "2" }

status.phase moves pending-approval → running → completed | failed; status.results[] records each action's outcome. See the remediation guide.