Overview
Pools are the logical containers for stored data. Each pool has a defined data protection policy (replication or erasure coding), a device class mapping, and a PG count. Creating pools correctly at the outset avoids disruptive reconfiguration later.Prerequisites
- Administrator credentials with the
adminrole - SSH access to a cluster management node
- CRUSH map configured with appropriate device class rules (see CRUSH Maps)
Creating Pools
- Replicated Pool
- Erasure-Coded Pool
Create the pool
Create a replicated pool
128 initially — the PG autoscaler will
adjust automatically as data volume grows.Set replication factor
Set replication size (factor)
size is the total number of copies. min_size is the minimum needed to
serve I/O (allows degraded operation with 2 copies during OSD failure recovery).Assign a CRUSH rule
Associate the pool with a CRUSH rule that targets the correct device class:Use
Set CRUSH rule on pool
ceph osd crush rule ls to list available rules.Enable pool application
Tag the pool with its application type so the cluster knows how to manage it:Valid application types:
Enable RBD (block storage) on pool
rbd (block), rgw (object), cephfs (file).Pool appears in
ceph osd pool ls detail with correct application and replication settings.Managing Existing Pools
- Inspect Pools
- Modify Pools
- Delete a Pool
List all pools with details
Show pool statistics
Show specific pool configuration
Pool Configuration Reference
Next Steps
CRUSH Maps
Configure failure domains and device class rules for pool placement
Storage Tiers
Map pools to Cinder volume types for multi-tier storage
Capacity Planning
Monitor pool utilization and plan expansion
Troubleshooting
Diagnose pool-related issues — PG warnings, capacity alerts