Appendix H
Changes in Reef, Squid and Tentacle
This appendix lists, one table per release, the notable changes in the three major releases Reef (18.2), Squid (19.2) and Tentacle (20.2) that matter for this book's chapters. Rows marked confirmed are stated in the official release notes of that release's x.2.0 version; rows marked for review come from the chapters or from point-release notes and have not yet been checked against the main release notes. For Umbrella (21.2) and each release's support dates, see appendix D.
Reef+ Reef (18.2)#
Version 18.2.0, 2023-08-07 — release notes: https://docs.ceph.com/en/latest/releases/reef/
| Area | Change | Chapters | Status |
|---|
| RADOS | FileStore is not supported; BlueStore is the only object store of the classic OSD (FileStore OSDs must be recreated before the upgrade). | Chapter 12, Chapter 10, Chapter 21 | confirmed |
| RADOS | Cache tiering is deprecated. | Chapter 16, Chapter 15 | confirmed |
| RADOS | RocksDB upgraded to 7.9.2, with RocksDB settings tunable per column family and new defaults. | Chapter 25, Chapter 12 | confirmed |
| RADOS | The read balancer: balancing primary PGs per pool, offline with osdmaptool and pg-upmap-primary. | Chapter 8, Chapter 19, Chapter 28 | confirmed |
| RADOS | mClock reworked: balanced becomes the default profile, QoS is a fraction of the OSD's IOPS capacity, and degraded objects recover before misplaced ones. | Chapter 21, Chapter 7, Chapter 10 | confirmed |
| RADOS | perf dump and perf schema are deprecated in favour of counter dump and counter schema (labelled perf counters). | Chapter 20, Chapter 10 | confirmed |
| Monitoring | ceph-exporter, deployed on every host, exports the daemons' performance metrics instead of the mgr's prometheus module. | Chapter 20 | confirmed |
| RGW | Bucket resharding works in multisite (the resharding zone feature, on by default for new deployments). | Chapter 5, Chapter 42, Chapter 40 | confirmed |
| RGW | Compression for objects uploaded with server-side encryption (the compress-encrypted zonegroup feature). | Chapter 5, Chapter 4 | confirmed |
| RBD | Layered client-side encryption: a clone can have its own encryption format and passphrase, distinct from its parent's. | Chapter 2, Chapter 38 | confirmed |
| Dashboard | A new Dashboard landing page with alert and chart cards, and a cephx user management page. | Chapter 9 | confirmed |
| RGW | The rgw mgr module automates multisite setup (ceph rgw realm bootstrap), as documented in the official Reef mgr/rgw module documentation. | Chapter 5 | confirmed |
Squid+ Squid (19.2)#
Version 19.2.0, 2024-09-26 — release notes: https://docs.ceph.com/en/latest/releases/squid/
| Area | Change | Chapters | Status |
|---|
| RADOS | LZ4 compression in BlueStore's RocksDB is on by default. | Chapter 12, Chapter 25 | confirmed |
| RADOS | A new CRUSH rule type, MSR (multi-step retry), for more flexible erasure-coded layouts. | Chapter 13, Chapter 28, Chapter 8 | confirmed |
| RADOS | The balancer module can manage read balancing online, with the new read and upmap-read modes. | Chapter 19, Chapter 28 | confirmed |
| RADOS | Scrub scheduling was improved. | Chapter 32, Chapter 17 | confirmed |
| RADOS | The POOL_APP_NOT_ENABLED warning is raised for every pool without an application tag, used or not. | Chapter 36, Chapter 7 | confirmed |
| Crimson | Crimson's first technology preview (RBD workloads on replicated pools). | Chapter 30 | confirmed |
| RBD | diff-iterate runs locally in fast-diff mode when the exclusive lock is available, a large speed-up for QEMU live disk sync and backups. | Chapter 2, Chapter 39 | confirmed |
| RGW | User accounts: AWS-compatible IAM APIs for self-service management of users, keys, groups, roles and policies. | Chapter 4, Chapter 41 | confirmed |
| RGW | A new layout for bucket-notification topics that replicates through metadata sync (the notification_v2 zone feature). | Chapter 5, Chapter 42 | confirmed |
| CephFS | Subvolume quiesce commands for crash-consistent snapshots, ceph fs swap, and --fs required by snap-schedule commands on multi-file-system clusters. | Chapter 6 | confirmed |
| cephx | Key rotation with ceph auth rotate. | Chapter 8 | confirmed |
| RBD | Point release 19.2.4 added the group argument of ceph orch apply nvmeof, so the NVMe-oF gateway with gateway groups is the official way to serve RBD over NVMe/TCP. | Chapter 2, Chapter 31 | confirmed |
| RADOS | Point release 19.2.1 changed the HDD OSD shard defaults to osd_op_num_shards_hdd 1 (was 5) and osd_op_num_threads_per_shard_hdd 5 (was 1), for steadier mClock client and recovery throughput (issue 66289); Tentacle and Umbrella keep it. | Chapter 21, Chapter 7 | confirmed |
Tentacle+ Tentacle (20.2)#
Version 20.2.0, 2025-11-18 — release notes: https://docs.ceph.com/en/latest/releases/tentacle/
| Area | Change | Chapters | Status |
|---|
| RADOS | FastEC: partial reads and writes and less space amplification in erasure-coded pools, enabled per pool with allow_ec_optimizations. | Chapter 13, Chapter 27, Chapter 16 | confirmed |
| RADOS | ISA-L replaces Jerasure as the default erasure-code plugin for clusters created on Tentacle; upgraded clusters keep their defaults. | Chapter 13, Chapter 27 | confirmed |
| RADOS | BlueStore: better compression and a new, faster write-ahead log. | Chapter 12, Chapter 24 | confirmed |
| RADOS | A per-pool data availability score (ceph osd pool availability-status, tech preview, off by default). | Chapter 20, Chapter 17 | confirmed |
| Crimson | SeaStore technology preview: deployable alongside the Crimson OSD for testing. | Chapter 30 | confirmed |
| MGR | The restful and zabbix mgr modules were removed; the dashboard module's REST API is the replacement for restful. | Chapter 7, Chapter 9, Chapter 10 | confirmed |
| cephadm | The mgmt-gateway service (one TLS entry point for the Dashboard and the monitoring stack), oauth2-proxy for SSO, and certmgr for certificates. | Chapter 9, Chapter 20 | confirmed |
| CephFS | The smb mgr module manages Samba-backed SMB shares on CephFS, with standalone users or Active Directory. | Chapter 6 | confirmed |
| CephFS | Case-insensitive or normalised directory entry names, and a confirmation flag for changing max_mds on an unhealthy cluster. | Chapter 6 | confirmed |
| RBD | rbd device map uses msgr2 by default (-o ms_mode=legacy keeps msgr1 for krbd). | Chapter 2, Chapter 29 | confirmed |
| RBD | Mirroring can pair a local namespace with a differently named remote one (--remote-namespace), and rbd group info and rbd group snap info were added. | Chapter 2, Chapter 39 | confirmed |
| RGW | S3 GetObjectAttributes; LastModified truncated to the second; resharding blocks writes for much less time. | Chapter 4, Chapter 40 | confirmed |
| RGW | Tenant-level IAM APIs are deprecated in favour of accounts (removal no sooner than the V release). | Chapter 4 | confirmed |
| RBD | Point release 20.2.1 added the ceph nvme-gw listeners command; the ceph nvmeof CLI for gateway configuration is documented in 20.2.5. | Chapter 31 | confirmed |
Umbrella+ Umbrella (21.2)#
Version 21.2.0, 2026-10-28 — release notes: https://docs.ceph.com/en/latest/releases/umbrella/
| Area | Change | Chapters | Status |
|---|
| cephadm | CRUSH-bucket-aware OSD upgrades: ok-to-upgrade now aggregates OSDs, and upgrades can be scoped to a CRUSH bucket. | Chapter 7 | confirmed |
Commands that need a release newer than Reef#
Release labels: every code block that runs a Ceph command carries a label such as Reef+, meaning the commands work unchanged on that release and later. The book's baseline is Reef; the blocks below need a newer release:
| Label | First command of the block | Chapter |
|---|
| Squid+ | ceph osd pool create nvmeof-pool && rbd pool init nvmeof-pool | Chapter 2 |
| Squid+ | radosgw-admin account create --account-name=acme --email=ops@acme.example | Chapter 4 |
| Tentacle+ | ceph mgr module enable smb | Chapter 6 |
| Umbrella+ | ceph orch upgrade start --image quay.io/ceph/ceph:v21.2.0 --daemon-types mgr,mon | Chapter 7 |
| Tentacle+ | ceph orch apply mgmt-gateway --placement="label:mgmt" | Chapter 9 |
| Tentacle+ | ceph osd pool set ecpool allow_ec_optimizations true | Chapter 13 |
| Tentacle+ | ceph osd erasure-code-profile set ec42fast plugin=isa technique=reed_sol_van k=4 m=2 \ | Chapter 13 |
| Tentacle+ | ceph orch apply mgmt-gateway --placement="label:mgmt" | Chapter 20 |
| Tentacle+ | ceph osd erasure-code-profile set ec42 plugin=isa technique=reed_sol_van k=4 m=2 crush-failure-domain=host | Chapter 27 |
| main | ceph tell osd.0 dump_metrics reactor_utilization | Chapter 30 |
| Tentacle+ | ceph orch apply nvmeof POOL GROUP --placement="host01, host02" | Chapter 31 |
| Squid+ | ceph pg deep-scrub 2.1f | Chapter 32 |