Learning Ceph(Second Edition)
上QQ阅读APP看书,第一时间看更新

Admin host

Management of Ceph clusters is usually done via a set of command-line interface (CLI) tools. While some Ceph managers perform these actions from one or more MON hosts, others choose to provision one or more dedicated, standalone servers for the purpose. These admin hosts require very few resources and are easily implemented as Virtual Machines (VMs) or even piggybacked onto bastion, jump, gateway, or other infrastructure services. One choice might be to share an existing Puppet or Ansible master host, which is likely set up in a conducive fashion. Be careful, though, to not provision Admin hosts with circular dependencies. If you use an OpenStack instance with a boot drive that lives on a Ceph cluster, sooner or later you'll find that cluster issues that require action prevent the admin host itself from working properly!

Another factor to consider when piggybacking the admin host role onto one or more Ceph MON nodes is that during cluster software upgrades and server hardware maintenance, each may in turn be rebooted or otherwise experience interrupted unavailability. While this does not impact cluster operations, it does prevent the continuity that is required of an admin host. For this reason as well, it is suggested to provision one or more admin hosts that are not hosted on or dependent on any other Ceph components.