ceph-cookbook

ceph-cookbook

A Claude Code plugin for deploying, operating, and troubleshooting Ceph storage clusters — skills, agents, hooks, and monitors covering cluster setup (Rook or cephadm), block/file/ object storage, day-2 lifecycle operations, security, and observability.

Installation

Add this repo as a plugin marketplace, then install the plugin:

/plugin marketplace add sserkanml/ceph-cookbook
/plugin install ceph-cookbook

Skills

Cluster deployment

Skill	Use for
ceph-deploy-rook	Bootstrap a Ceph cluster on Kubernetes via the Rook operator + CephCluster CRD
ceph-deploy-cephadm	Bootstrap a Ceph cluster on bare-metal/VMs via cephadm + ceph orch

Object storage (RGW)

Skill	Use for
rgw-multisite-ceph	RGW multisite replication between two cephadm-managed clusters (radosgw-admin + ceph orch)
rgw-multisite-rook	RGW multisite replication between two Rook-managed clusters (CephObjectRealm/CephObjectZone CRDs)
rgw-user-quota-mgmt	RGW S3 user lifecycle, access key rotation, quotas, bucket policies

Block & file storage

Skill	Use for
rbd-pool-setup	Create an RBD pool/image, scope a cephx client, map/mount
rbd-mirror-dr	RBD mirroring (journal or snapshot mode) between two clusters for block-storage DR
cephfs-deploy	Deploy MDS daemons and create a CephFS filesystem via cephadm
nfs-ganesha-export	Export CephFS or an RGW bucket over NFS via ceph orch-managed NFS-Ganesha

Cluster lifecycle & tuning

Skill	Use for
ceph-upgrade	Rolling cluster upgrade via ceph orch upgrade, with health gating
osd-host-lifecycle	Add/drain/remove OSDs and hosts, replace a failed disk
crush-pool-tuning	Replicated vs. erasure-coded pools, CRUSH failure domains, PG autoscaling

Security & observability

Skill	Use for
cephx-key-management	Least-privilege client keyrings, capability auditing, key rotation/revocation
ceph-prometheus-grafana	Enable the mgr prometheus module and wire up Prometheus/Grafana dashboards

Agents

Agent	Use for
ceph-diagnostician	Read-only health/OSD/PG/RGW-sync diagnosis; reports issues and recommended next steps, never remediates
ceph-remediator	Executes a fix for an already-diagnosed issue, with explicit risk classification and confirmation before medium/high-risk actions

Hooks & Monitors

• guard-dangerous-commands.sh (PreToolUse hook) — blocks irreversible commands (pool delete, OSD purge, realm/zone delete, force-repair, --yes-i-really-mean-it) before they run, regardless of prior confirmation.
• health-watch.sh (monitor) — reports HEALTH_OK/WARN/ERR transitions.
• rgw-sync-watch.sh (monitor) — reports RGW multisite sync status transitions (caught up / behind / failed).

Output Styles

Style	Use for
ceph-incident	Fast, command-first responses during a live incident
ceph-postmortem	Root-cause analysis and prevention, after the fact
ceph-runbook	Step-by-step, verifiable procedures for planned maintenance/setup

License

GPL-3.0 — see LICENSE.