Cephadm 概念

全资域名 vs 裸主机名

cephadm has very minimal requirements when it comes to resolving host names etc. When cephadm initiates an ssh connection to a remote host, the host name can be resolved in four different ways:

  • a custom ssh config resolving the name to an IP

  • via an externally maintained /etc/hosts

  • via explictly providing an IP address to cephadm: ceph orch host add <hostname> <IP>

  • automatic name resolution via DNS.

Ceph itself uses the command hostname to determine the name of the current host.

Note

cephadm demands that the name of the host given via ceph orch host add equals the output of hostname on remote hosts.

Otherwise cephadm can’t be sure, the host names returned by ceph * metadata match the hosts known to cephadm. This might result in a CEPHADM_STRAY_HOST warning.

When configuring new hosts, there are two valid ways to set the hostname of a host:

  1. Using the bare host name. In this case:

  • hostname returns the bare host name.

  • hostname -f returns the FQDN.

  1. Using the fully qualified domain name as the host name. In this case:

  • hostname returns the FQDN

  • hostname -s return the bare host name

Note that man hostname recommends hostname to return the bare host name:

The FQDN (Fully Qualified Domain Name) of the system is the name that the resolver(3) returns for the host name, such as, ursula.example.com. It is usually the hostname followed by the DNS domain name (the part after the first dot). You can check the FQDN using hostname –fqdn or the domain name using dnsdomainname.

You cannot change the FQDN with hostname or dnsdomainname.

The recommended method of setting the FQDN is to make the hostname
be an alias for the fully qualified name using /etc/hosts, DNS, or
NIS. For example, if the hostname was "ursula", one might have
a line in /etc/hosts which reads

       127.0.1.1    ursula.example.com ursula

Which means, man hostname recommends hostname to return the bare host name. This in turn means that Ceph will return the bare host names when executing ceph * metadata. This in turn means cephadm also requires the bare host name when adding a host to the cluster: ceph orch host add <bare-name>.

Cephadm Scheduler

Cephadm uses a declarative state to define the layout of the cluster. This state consists of a list of service specificatins containing placement specifications (See Service Specification ).

Cephadm constantly compares list of actually running daemons in the cluster with the desired service specifications and will either add or remove new daemons.

First, cephadm will select a list of candidate hosts. It first looks for explicit host names and will select those. In case there are no explicit hosts defined, cephadm looks for a label specification. If there is no label defined in the specification, cephadm will select hosts based on a host pattern. If there is no pattern defined, cepham will finally select all known hosts as candidates.

Then, cephadm will consider existing daemons of this servics and will try to avoid moving any daemons.

Cephadm supports the deployment of a specific amount of services. Let’s consider a service specification like so:

service_type: mds
service_name: myfs
placement:
  count: 3
  label: myfs

This instructs cephadm to deploy three daemons on hosts labeld with myfs across the cluster.

Then, in case there are less than three daemons deployed on the candidate hosts, cephadm will then then randomly choose hosts for deploying new daemons.

In case there are more than three daemons deployed, cephadm will remove existing daemons.

Finally, cephadm will remove daemons on hosts that are outside of the list of candidate hosts.

However, there is a special cases that cephadm needs to consider.

In case the are fewer hosts selected by the placement specification than demanded by count, cephadm will only deploy on selected hosts.