ceph-deploy man page

ceph-deploy — Ceph deployment tool


ceph-deploy new [initial-monitor-node(s)]
ceph-deploy install [ceph-node] [ceph-node...]
ceph-deploy mon create-initial
ceph-deploy osd prepare [ceph-node]:[dir-path]
ceph-deploy osd activate [ceph-node]:[dir-path]
ceph-deploy osd create [ceph-node]:[dir-path]
ceph-deploy admin [admin-node][ceph-node...]
ceph-deploy purgedata [ceph-node][ceph-node...]
ceph-deploy forgetkeys


ceph-deploy is a tool which allows easy and quick deployment of a Ceph cluster without involving complex and detailed manual configuration. It uses ssh to gain access to other Ceph nodes from the admin node, sudo for administrator privileges on them and the underlying Python scripts automates the manual process of Ceph installation on each node from the admin node itself. It can be easily run on an workstation and doesn't require servers, databases or any other automated tools. With ceph-deploy, it is really easy to set up and take down a cluster. However, it is not a generic deployment tool. It is a specific tool which is designed for those who want to get Ceph up and running quickly with only the unavoidable initial configuration settings and without the overhead of installing other tools like Chef, Puppet or Juju. Those who want to customize security settings, partitions or directory locations and want to set up a cluster following detailed manual steps, should use other tools i.e, Chef, Puppet, Juju or Crowbar.

With ceph-deploy, you can install Ceph packages on remote nodes, create a cluster, add monitors, gather/forget keys, add OSDs and metadata servers, configure admin hosts or take down the cluster.



Start deploying a new cluster and write a configuration file and keyring for it. It tries to copy ssh keys from admin node to gain passwordless ssh to monitor node(s), validates host IP, creates a cluster with a new initial monitor node or nodes for monitor quorum, a ceph configuration file, a monitor secret keyring and a log file for the new cluster. It populates the newly created Ceph configuration file with fsid of cluster, hostnames and IP addresses of initial monitor members under [global] section.


ceph-deploy new [MON][MON...]

Here, [Mon] is the initial monitor hostname (short hostname i.e, hostname -s).

Other options like --no-ssh-copykey, --fsid, --cluster-network and --public-network can also be used with this command.

If more than one network interface is used, public network setting has to be added under [global] section of Ceph configuration file. If the public subnet is given, new command will choose the one IP from the remote host that exists within the subnet range. Public network can also be added at runtime using --public-network option with the command as mentioned above.


Install Ceph packages on remote hosts. As a first step it installs yum-plugin-priorities in admin and other nodes using passwordless ssh and sudo so that Ceph packages from upstream repository get more priority. It then detects the platform and distribution for the hosts and installs Ceph normally by downloading distro compatible packages if adequate repo for Ceph is already added. --release flag is used to get the latest release for installation. During detection of platform and distribution before installation, if it finds the distro.init to be sysvinit (Fedora, CentOS/RHEL etc), it doesn't allow installation with custom cluster name and uses the default name ceph for the cluster.

If the user explicitly specifies a custom repo url with --repo-url for installation, anything detected from the configuration will be overridden and the custom repository location will be used for installation of Ceph packages. If required, valid custom repositories are also detected and installed. In case of installation from a custom repo a boolean is used to determine the logic needed to proceed with a custom repo installation. A custom repo install helper is used that goes through config checks to retrieve repos (and any extra repos defined) and installs them. cd_conf is the object built from argparse that holds the flags and information needed to determine what metadata from the configuration is to be used.

A user can also opt to install only the repository without installing Ceph and its dependencies by using --repo option.


ceph-deploy install [HOST][HOST...]

Here, [HOST] is/are the host node(s) where Ceph is to be installed.

An option --release is used to install a release known as CODENAME (default: firefly).

Other options like --testing, --dev, --adjust-repos, --no-adjust-repos, --repo, --local-mirror, --repo-url and --gpg-url can also be used with this command.


Deploy Ceph mds on remote hosts. A metadata server is needed to use CephFS and the mds command is used to create one on the desired host node. It uses the subcommand create to do so. create first gets the hostname and distro information of the desired mds host. It then tries to read the bootstrap-mds key for the cluster and deploy it in the desired host. The key generally has a format of {cluster}.bootstrap-mds.keyring. If it doesn't finds a keyring, it runs gatherkeys to get the keyring. It then creates a mds on the desired host under the path /var/lib/ceph/mds/ in /var/lib/ceph/mds/{cluster}-{name} format and a bootstrap keyring under /var/lib/ceph/bootstrap-mds/ in /var/lib/ceph/bootstrap-mds/{cluster}.keyring format. It then runs appropriate commands based on distro.init to start the mds. To remove the mds, subcommand destroy is used.


ceph-deploy mds create [HOST[:DAEMON-NAME]] [HOST[:DAEMON-NAME]...]

ceph-deploy mds destroy [HOST[:DAEMON-NAME]] [HOST[:DAEMON-NAME]...]

The [DAEMON-NAME] is optional.


Deploy Ceph monitor on remote hosts. mon makes use of certain subcommands to deploy Ceph monitors on other nodes.

Subcommand create-initial deploys for monitors defined in mon initial members under [global] section in Ceph configuration file, wait until they form quorum and then gatherkeys, reporting the monitor status along the process. If monitors don't form quorum the command will eventually time out.


ceph-deploy mon create-initial

Subcommand create is used to deploy Ceph monitors by explicitly specifying the hosts which are desired to be made monitors. If no hosts are specified it will default to use the mon initial members defined under [global] section of Ceph configuration file. create first detects platform and distro for desired hosts and checks if hostname is compatible for deployment. It then uses the monitor keyring initially created using new command and deploys the monitor in desired host. If multiple hosts were specified during new command i.e, if there are multiple hosts in mon initial members and multiple keyrings were created then a concatenated keyring is used for deployment of monitors. In this process a keyring parser is used which looks for [entity] sections in monitor keyrings and returns a list of those sections. A helper is then used to collect all keyrings into a single blob that will be used to inject it to monitors with --mkfs on remote nodes. All keyring files are concatenated to be in a directory ending with .keyring. During this process the helper uses list of sections returned by keyring parser to check if an entity is already present in a keyring and if not, adds it. The concatenated keyring is used for deployment of monitors to desired multiple hosts.


ceph-deploy mon create [HOST] [HOST...]

Here, [HOST] is hostname of desired monitor host(s).

Subcommand add is used to add a monitor to an existing cluster. It first detects platform and distro for desired host and checks if hostname is compatible for deployment. It then uses the monitor keyring, ensures configuration for new monitor host and adds the monitor to the cluster. If the section for the monitor exists and defines a mon addr that will be used, otherwise it will fallback by resolving the hostname to an IP. If --address is used it will override all other options. After adding the monitor to the cluster, it gives it some time to start. It then looks for any monitor errors and checks monitor status. Monitor errors arise if the monitor is not added in mon initial members, if it doesn't exist in monmap and if neither public_addr nor public_network keys were defined for monitors. Under such conditions, monitors may not be able to form quorum. Monitor status tells if the monitor is up and running normally. The status is checked by running ceph daemon mon.hostname mon_status on remote end which provides the output and returns a boolean status of what is going on. False means a monitor that is not fine even if it is up and running, while True means the monitor is up and running correctly.


ceph-deploy mon add [HOST]

ceph-deploy mon add [HOST] --address [IP]

Here, [HOST] is the hostname and [IP] is the IP address of the desired monitor node. Please note, unlike other mon subcommands, only one node can be specified at a time.

Subcommand destroy is used to completely remove monitors on remote hosts. It takes hostnames as arguments. It stops the monitor, verifies if ceph-mon daemon really stopped, creates an archive directory mon-remove under /var/lib/ceph/, archives old monitor directory in {cluster}-{hostname}-{stamp} format in it and removes the monitor from cluster by running ceph remove... command.


ceph-deploy mon destroy [HOST] [HOST...]

Here, [HOST] is hostname of monitor that is to be removed.


Gather authentication keys for provisioning new nodes. It takes hostnames as arguments. It checks for and fetches client.admin keyring, monitor keyring and bootstrap-mds/bootstrap-osd keyring from monitor host. These authentication keys are used when new monitors/OSDs/Mds are added to the cluster.


ceph-deploy gatherkeys [HOST] [HOST...]

Here, [HOST] is hostname of the monitor from where keys are to be pulled.


Manage disks on a remote host. It actually triggers the ceph-disk utility and it's subcommands to manage disks.

Subcommand list lists disk partitions and Ceph OSDs.


ceph-deploy disk list [HOST:[DISK]]

Here, [HOST] is hostname of the node and [Disk] is disk name or path.

Subcommand prepare prepares a directory, disk or drive for a Ceph Osd. It creates a GPT partition, marks the partition with Ceph type uuid, creates a file system, marks the file system as ready for Ceph consumption, uses entire partition and adds a new partition to the journal disk.


ceph-deploy disk prepare [HOST:[DISK]]

Here, [HOST] is hostname of the node and [Disk] is disk name or path.

Subcommand activate activates the Ceph Osd. It mounts the volume in a temporary location, allocates an Osd id (if needed), remounts in the correct location /var/lib/ceph/osd/$cluster-$id and starts ceph-osd. It is triggered by udev when it sees the Osd GPT partition type or on ceph service start with ceph disk activate-all.


ceph-deploy disk activate [HOST:[DISK]]

Here, [HOST] is hostname of the node and [Disk] is disk name or path.

Subcommand zap zaps/erases/destroys a device's partition table and contents. It actually uses sgdisk and it's option --zap-all to destroy both GPT and MBR data structures so that the disk becomes suitable for repartitioning. sgdisk then uses --mbrtogpt to convert the MBR or BSD disklabel disk to a GPT disk. The prepare subcommand can now be executed which will create a new GPT partition.


ceph-deploy disk zap [HOST:[DISK]]

Here, [HOST] is hostname of the node and [Disk] is disk name or path.


Manage OSDs by preparing data disk on remote host. osd makes use of certain subcommands for managing OSDs.

Subcommand prepare prepares a directory, disk or drive for a Ceph Osd. It first checks against multiple OSDs getting created and warns about the possibility of more than the recommended which would cause issues with max allowed PIDs in a system. It then reads the bootstrap-osd key for the cluster or writes the bootstrap key if not found. It then uses ceph-disk utility's prepare subcommand to prepare the disk, journal and deploy the Osd on the desired host. Once prepared, it gives some time to the Osd to settle and checks for any possible errors and if found, reports to the user.


ceph-deploy osd prepare HOST:DISK[:JOURNAL] [HOST:DISK[:JOURNAL]...]

Subcommand activate activates the Osd prepared using prepare subcommand. It actually uses ceph-disk utility's activate subcommand with appropriate init type based on distro to activate the Osd. Once activated, it gives some time to the Osd to start and checks for any possible errors and if found, reports to the user. It checks the status of the prepared Osd, checks the Osd tree and makes sure the OSDs are up and in.


ceph-deploy osd activate HOST:DISK[:JOURNAL] [HOST:DISK[:JOURNAL]...]

Subcommand create uses prepare and activate subcommands to create an Osd.


ceph-deploy osd create HOST:DISK[:JOURNAL] [HOST:DISK[:JOURNAL]...]

Subcommand list lists disk partitions, Ceph OSDs and prints Osd metadata. It gets the osd tree from a monitor host, uses the ceph-disk-list output and gets the mount point by matching the line where the partition mentions the Osd name, reads metadata from files, checks if a journal path exists, if the Osd is in a Osd tree and prints the Osd metadata.


ceph-deploy osd list HOST:DISK[:JOURNAL] [HOST:DISK[:JOURNAL]...]


Push configuration and client.admin key to a remote host. It takes the {cluster}.client.admin.keyring from admin node and writes it under /etc/ceph directory of desired node.


ceph-deploy admin [HOST] [HOST...]

Here, [HOST] is desired host to be configured for Ceph administration.


Push/pull configuration file to/from a remote host. It uses push subcommand to takes the configuration file from admin host and write it to remote host under /etc/ceph directory. It uses pull subcommand to do the opposite i.e, pull the configuration file under /etc/ceph directory of remote host to admin node.


ceph-deploy push [HOST] [HOST...]

ceph-deploy pull [HOST] [HOST...]

Here, [HOST] is the hostname of the node where config file will be pushed to or pulled from.


Remove Ceph packages from remote hosts. It detects the platform and distro of selected host and uninstalls Ceph packages from it. However, some dependencies like librbd1 and librados2 will not be removed because they can cause issues with qemu-kvm.


ceph-deploy uninstall [HOST] [HOST...]

Here, [HOST] is hostname of the node from where Ceph will be uninstalled.


Remove Ceph packages from remote hosts and purge all data. It detects the platform and distro of selected host, uninstalls Ceph packages and purges all data. However, some dependencies like librbd1 and librados2 will not be removed because they can cause issues with qemu-kvm.


ceph-deploy purge [HOST] [HOST...]

Here, [HOST] is hostname of the node from where Ceph will be purged.


Purge (delete, destroy, discard, shred) any Ceph data from /var/lib/ceph. Once it detects the platform and distro of desired host, it first checks if Ceph is still installed on the selected host and if installed, it won't purge data from it. If Ceph is already uninstalled from the host, it tries to remove the contents of /var/lib/ceph. If it fails then probably OSDs are still mounted and needs to be unmounted to continue. It unmount the OSDs and tries to remove the contents of /var/lib/ceph again and checks for errors. It also removes contents of /etc/ceph. Once all steps are successfully completed, all the Ceph data from the selected host are removed.


ceph-deploy purgedata [HOST] [HOST...]

Here, [HOST] is hostname of the node from where Ceph data will be purged.


Remove authentication keys from the local directory. It removes all the authentication keys i.e, monitor keyring, client.admin keyring, bootstrap-osd and bootstrap-mds keyring from the node.


ceph-deploy forgetkeys


Manage packages on remote hosts. It is used for installing or removing packages from remote hosts. The package names for installation or removal are to be specified after the command. Two options --install and --remove are used for this purpose.


ceph-deploy pkg --install [PKGs] [HOST] [HOST...]

ceph-deploy pkg --remove [PKGs] [HOST] [HOST...]

Here, [PKGs] is comma-separated package names and [HOST] is hostname of the remote node where packages are to be installed or removed from.


Install and configure Calamari nodes. It first checks if distro is supported for Calamari installation by ceph-deploy. An argument connect is used for installation and configuration. It checks for ceph-deploy configuration file (cd_conf) and Calamari release repo or calamari-minion repo. It relies on default for repo installation as it doesn't install Ceph unless specified otherwise. options dictionary is also defined because ceph-deploy pops items internally which causes issues when those items are needed to be available for every host. If the distro is Debian/Ubuntu, it is ensured that proxy is disabled for calamari-minion repo. calamari-minion package is then installed and custom repository files are added. minion config is placed prior to installation so that it is present when the minion first starts. config directory, calamari salt config are created and salt-minion package is installed. If the distro is Redhat/CentOS, the salt-minion service needs to be started.


ceph-deploy calamari {connect} [HOST] [HOST...]

Here, [HOST] is the hostname where Calamari is to be installed.

An option --release can be used to use a given release from repositories defined in ceph-deploy's configuration. Defaults to calamari-minion.

Another option --master can also be used with this command.


The current installed version of ceph-deploy.
The username to connect to the remote host.
Overwrite an existing conf file on remote host (if present).
Name of the cluster.
Use (or reuse) a given ceph.conf file.
Do not attempt to copy ssh keys.
Provide an alternate FSID for ceph.conf generation.
Specify the (internal) cluster network.
Specify the public network for a cluster.
Install the latest development release.
Install a bleeding edge built from Git branch or tag (default: master).
Install packages modifying source repos.
Install packages without modifying source repos.
Install repo files only (skips package installation).
Fetch packages and push them to hosts for a local repo mirror.
Specify a repo url that mirrors/contains Ceph packages.
Specify a GPG key url to be used with custom repos (defaults to ceph.com).
IP address of the host node to be added to the cluster.
Concatenate multiple keyrings to be seeded on new monitors.
Destroy the partition table and content of a disk.
Filesystem to use to format disk (e.g., xfs, btrfs).
Encrypt [data-path] and/or journal devices with dm-crypt.
Directory where dm-crypt keys are stored.
Comma-separated package(s) to install on remote hosts.
Comma-separated package(s) to remove from remote hosts.
The domain for the Calamari master server.


ceph-deploy is part of Ceph, a massively scalable, open-source, distributed storage system. Please refer to the documentation at http://ceph.com/ceph-deploy/docs for more information.

See Also

ceph-mon(8), ceph-osd(8), ceph-disk(8), ceph-mds(8)

Referenced By

ceph-detect-init(8), ceph-disk(8).

Explore man page connections for ceph-deploy(8).

Ceph Sep 29, 2016