The Ceph Blog

Ceph blog stories provide high-level spotlights on our customers all over the world

  • July 24, 2019
    New in Nautilus: Orchestrator

    Ceph Nautilus introduces a new orchestrator interface that provides the ability to control external deployment tools like ceph-ansible, DeepSea or Rook.  The vision is to provide a bridge between administrators, Ceph, and external deployment systems. In order to accomplish that, the orchestrator interface enables the Ceph dashboard or the ceph command line tool to access to data provided by different deployment tools …Read more

  • April 30, 2019
    New in Nautilus: crash dump telemetry

    When Ceph daemons encounter software bugs, unexpected state, failed assertions, or other exceptional cases, they dump a stack trace and recently internal log activity to their log file in /var/log/ceph. On modern systems, systemd will restart the daemon and life will go on–often without the cluster administrator even realizing that there was a problem. This …Read more

  • April 18, 2019
    New in Nautilus: ceph-iscsi Improvements

    The ceph-iscsi project provides a framework, REST API and CLI tool for creating and managing iSCSI targets and gateways for Ceph via LIO. It is the successor and a consolidation of two formerly separate projects, the ceph-iscsi-cli and ceph-iscsi-config which were initially started in 2016 by Paul Cuzner at Red Hat. While this is not a new feature of Ceph Nautilus per se, improving …Read more

  • April 12, 2019
    New in Nautilus: device management and failure prediction

    Ceph storage clusters ultimately rely on physical hardware devices–HDDs or SSDs–that can fail. Starting in Nautilus, management and tracking of physical devices is now handled by Ceph. Furthermore, we’ve added infrastructure to collect device health metrics (e.g., SMART) and to predict device failures before they happen, either via a built-in pre-trained prediction model, or via …Read more

  • April 4, 2019
    New in Nautilus: PG merging and autotuning

    Since the beginning, choosing and tuning the PG count in Ceph has been one of the more frustrating parts of managing a Ceph cluster.  Guidance for choosing an appropriate pool is confusing, inconsistent between sources, and frequently surrounded by caveats and exceptions.  And most importantly, if a bad value is chosen, it can’t always be …Read more