Ceph (简体中文)

From ArchWiki
翻译状态：本文是 Ceph 的翻译。上次翻译日期：2017-12-20。如果英文版本有所更改，则您可以帮助同步翻译。

Ceph 是一个专注于分布式的、弹性可扩展的、高可靠的、性能优异的存储系统平台，可用于为虚拟机提供块存储方案或通过 FUSE 提供常规的文件系统。Ceph 是个高度可配置的系统，管理者可以控制系统的各个方面。它提供了一个命令行界面用于监视和控制其存储集群。Ceph 也包含鉴证和授权功能，可兼容多种存储网关接口，如 OpenStack Swift 和 Amazon S3。

引自 Wikipedia: Ceph (software):

Ceph is a free software storage platform designed to present object, block, and file storage from a single distributed computer cluster. Ceph's main goals are to be completely distributed without a single point of failure, scalable to the exabyte level, and freely-available. The data is replicated, making it fault tolerant.

引自 Ceph.com:

Ceph is a distributed object store and file system designed to provide excellent performance, reliability and scalability.
警告： 推荐使用官方部署工具安装 Ceph 。该工具通过 SSH 连接到目标机器并自动完成安装、配置和系统管理。官方部署工具(ceph-deploy)目前尚不支持 Arch Linux ，不能使用快速安装方式[失效链接 2020-12-20 ⓘ] 部署，只能按官方文档手工部署[失效链接 2020-12-20 ⓘ]。因此本文目前仅介绍手工部署方法。

The official documentation states[失效链接 2020-12-20 ⓘ] "the manual procedure is primarily for exemplary purposes for those developing deployment scripts with Chef, Juju, Puppet, etc.".

术语

提示： 官方文档[失效链接 2020-12-20 ⓘ]提供了完整的术语表
  • Client : Something which connects to a Ceph cluster to access data but is not part of the Ceph cluster itself.
  • MONs : Also known as monitors, these store cluster state and maps containing information about the cluster such as running services and data locations.
  • MDSs : Also known as metadata servers, these store metadata for the Ceph filesystem to reduce load on the storage cluster (e.g. information for commands such as ls).
  • Node : A machine which is running Ceph services, such as OSDs or MONs.
  • OSDs : Also known as OSD daemons, these are responsible for the storage of data within the cluster and also conduct various related operations such as replication, recovery, and rebalancing.
  • Storage cluster : The core set of software responsible for storing data (OSDs+MONs).

安装

软件包

可以从官方源安装ceph。如果愿意冒险，也可以安装开发版的 ceph-gitAUR。

存储集群的所有节点都要安装 ceph。

NTP 客户端

警告： 应当同步监视器节点的时钟以避免时钟偏移，否则将导致集群性能下降甚至停止工作。官方文档[失效链接 2020-12-20 ⓘ]建议所有节点都应采取某种方式同步时钟。

在节点上安装并运行时钟同步客户端，可参阅时钟同步。

启动一个存储集群

Before a storage cluster can operate, the monitors for that cluster must be bootstrapped with several identifiers and keyrings.

The upstream Ceph documentation is well-written and kept updated with the latest releases.

To boostrap a storage cluster, follow the steps documented in the official manual deployment guide[失效链接 2020-12-20 ⓘ].

启动一个监视器

Since your system most likely uses systemd, you can enable a monitor as a systemd unit.

As an example, for a monitor named node1 start and enable [email protected] as detailed in Systemd#Using units.

参阅