AgentSkillsCN

storage-preflight

在部署前检查存储与 CSI 的健康状况。适用于部署应用、创建 PVC、添加需要持久化存储的工作负载，或当用户提及 Helm、部署或安装且带有存储需求时使用。

SKILL.md
--- frontmatter
name: storage-preflight
description: Check storage and CSI health before deployments. Use when deploying apps, creating PVCs, adding workloads that need persistent storage, or when user mentions helm, deploy, or install with storage requirements.
user-invocable: false

Storage Preflight Check

Automatically validates storage subsystem health before deploying workloads that require persistent storage.

When This Activates

  • •User is deploying a new app with PVCs
  • •User mentions adding storage to a workload
  • •User is installing a Helm chart that typically needs storage
  • •User creates or modifies PersistentVolumeClaims

Preflight Checks

1. CSI Driver Registration

bash
# Verify CSI drivers are registered
kubectl get csidrivers 2>/dev/null | grep -E 'ceph|rbd|cephfs'

Expected: storage.rbd.csi.ceph.com and storage.cephfs.csi.ceph.com present.

2. CSI Nodeplugin Health

bash
# Check nodeplugin pods are running (not crash-looping)
kubectl get pods -n storage -l app=csi-rbdplugin -o wide 2>/dev/null
kubectl get pods -n storage -l app=csi-cephfsplugin -o wide 2>/dev/null

# Check for high restart counts (>10 suggests instability)
kubectl get pods -n storage -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.status.containerStatuses[0].restartCount}{"\n"}{end}' 2>/dev/null | awk '$2 > 10'

3. Ceph Cluster Health

bash
# Quick Ceph health check
kubectl -n storage exec deploy/rook-ceph-tools -- ceph health 2>/dev/null

Must be HEALTH_OK or HEALTH_WARN (not HEALTH_ERR).

4. Storage Class Availability

bash
# Verify default storage class exists
kubectl get storageclass | grep -E 'ceph-block|ceph-filesystem'

5. Pending PVCs (indicates existing issues)

bash
# Check for stuck PVCs
kubectl get pvc -A --field-selector=status.phase=Pending 2>/dev/null

Any pending PVCs suggest storage provisioning issues.

Output Format

If all checks pass:

code
✅ Storage preflight passed - safe to deploy

If issues found:

code
⚠️ Storage preflight warnings:

**Issue**: [description]
**Impact**: [what will happen if you proceed]
**Fix**: [how to resolve]

Recommend fixing storage issues before deploying.

Blocking Issues (STOP deployment)

  • •CSI drivers not registered
  • •Ceph HEALTH_ERR
  • •CSI nodeplugins in CrashLoopBackOff
  • •Multiple pending PVCs

Warning Issues (proceed with caution)

  • •High restart counts on CSI pods
  • •Ceph HEALTH_WARN
  • •Single pending PVC (might be unrelated)

Skip Conditions

Don't run this check if:

  • •Deployment explicitly has no PVCs/storage
  • •User is deploying to a namespace that never uses Ceph
  • •User explicitly says "skip storage check"