What’s New in Ceph? Highlights from the 2018 Developer Monthly Report

This article is the sixth installment of the 2018 Ceph Developer Monthly Report, where the UMCloud storage team shares interesting commits and important changes from the Ceph community each month.

Three main highlights this issue:

ceph‑disk is deprecated; the next release will require ceph‑volume.

seastar‑based OSD is on the way.

Disk‑failure prediction based on SMART test results is arriving.

Block Storage

rbd adds basic namespace support

Pull request #22673 : rbd: basic support for images within namespaces.

Pull request #22608 : librbd: namespace create/remove/list support.

Unified Storage Layer

BlueStore allocator reduces memory usage

BlueStore introduces a new bitmap allocator for free‑space management, now the default for both bluefs and BlueStore.

Pull request #21825 : os/bluestore: new bitmap allocator.

Cluster Management

ceph‑disk removed; use ceph‑volume

Pull request #22343 : ceph‑disk: removal.

New osd destroy‑new command retains OSD ID while deleting data

Pull request #22429 : mon: add ‘osd destroy‑new’ command that only destroys NEW OSD slots.

Collect SMART information and persist to a dedicated pool

Pull request #22479 : mgr/devicehealth: devicehealth module base – first step for native SMART‑based disk‑failure prediction.

Foundations Library

Network message layer adds iWARP RDMA protocol support

Pull request #20297 : msg/async/rdma: add iWARP RDMA protocol support.

OSD crash metadata dump to /var/lib/ceph/crash

Pull request #22586 : common: dump crash metadata to /var/lib/ceph/crash.

Progress on seastar‑based OSD rewrite

The work is divided into three parts:

Network message layer supports seastar – pull request #22491 .

Memory management supports seastar – pull request #22454 .

New thread pool implementation based on seastar – pull request #22565 .

This Month’s Commit Statistics

— END —