Ceph

Tentacle (v20) ships a major feature surface; latest point release 20.2.2 (2026-06-16). The v20 line went GA as 20.2.0 on 2025-11-18, with 20.2.1 (2026-04-06) and 20.2.2 (2026-06-16) following. It is actively maintained alongside Squid (v19). Per Ceph Tentacle release notes.

FastEC — erasure-coding gets partial reads/writes. Tentacle introduces long-anticipated performance and space-amplification optimizations for erasure-coded pools, including partial reads and partial writes, with a new ISA-L plugin now the default for new EC profiles. This materially changes EC economics vs. 3x replication. Per Ceph Tentacle release notes.

Integrated SMB and NVMe/TCP gateway. A new SMB Manager module deploys Samba-backed file shares over CephFS (domain and standalone auth); the NVMe/TCP gateway adds gateway groups, multiple namespaces, DHCHAP keys, and namespace encryption. Crimson/SeaStore is now deployable as a tech preview toward replacing the classical OSD. Per Ceph Tentacle release notes.

Dashboard modernization + management gateway. Tentacle ships a redesigned dashboard with OAuth 2.0 integration and multi-cluster management, plus a new cephadm nginx-based management gateway providing TLS-terminated, HA access to management endpoints with optional OAuth2 single sign-on. RBD live migration can now instantly import images from other clusters via NBD streams. Per Ceph Tentacle release notes.

Community substrate rebuilt around regional Ceph Days. The Foundation shifted from one global mega-event toward regional Ceph Days — India drew 192 attendees, Raleigh 78 — and restarted Ceph Tech Talks after a long gap, with documentation the explicit Q1 priority. Per Ceph Q1 2026 Newsletter.

Operational signals from the field. Two recurring 2026 failure modes: a 5-node Ceph 19.2.3 cluster freezing on node reentry under STP ring-topology peering storms (reported on the Proxmox forum), and CephFS outages after Debian Trixie upgrade/reboot cycles where the MDS fails to rejoin cleanly. Community runbooks now lean harder on total-power-loss recovery — worth pre-rehearsing in production. Per Proxmox forum.

MinIO's retreat from open source is pushing evaluators back toward Ceph. With MinIO curtailing its open-source offering, 2026 comparison guides now frame upstream Ceph as the mature, production-ready alternative — though the real cost of either option is operational knowledge, not licensing. Per Ceph vs MinIO 2026 storage guide.

The 2026 DRAM/NAND price spike hits Ceph clusters harder than most. DRAM prices rose roughly 105% in Q1 2026 (TrendForce forecast a further 58-63% rise), and enterprise SSDs rose about 60% — a cost pressure that lands unusually hard on Ceph specifically, since each OSD carries a 1-2GB RAM baseline that multiplies across large clusters. Per The Memory Crisis Is Making Your Ceph Deployment More Expensive.

A NIC firmware bug took down S3-compatible storage during Ceph backfill traffic. An April 30 2026 incident traced a six-hour outage to an Intel 10GbE NIC running old firmware (1.3429.0) hitting a Tx Unit Hang under sustained Ceph backfill load; resolved by physical NIC replacement, no data loss. Per Incident review — S3 object storage instability, April 30 2026.

Ceph Foundation is a Linux Foundation directed fund, not a single-vendor operation. Established in 2018, the Foundation exists to provide "responsible and beneficial support, independent of the technical direction of Ceph" — vendor-neutral governance that's easy to miss for a project so often framed purely through its Red Hat/IBM lineage, and a structural fact this node doesn't currently state despite covering Ceph's community substrate (Ceph Days, Tech Talks) in detail. Per Ceph Foundation. Sources: Ceph Tentacle release notes · Ceph Q1 2026 Newsletter · Proxmox forum