Sysadmin Garden of Eden

Ceph: Build your own cluster on CentOS

Table of Contents
  1. 1. Requirements
  2. 2. Preparing your servers for Ceph
  3. 3. Preparing the Ceph deployment
  4. 4. Deploying Ceph storage nodes
  5. 5. Creating the storage nodes
  6. 6. Finalizing your Ceph cluster
  7. 7. Troubleshooting
    1. 7.1. Log file location
    2. 7.2. “Clock skew” error

Ceph is an object, a block and filesystem storage system.

As you can see from the Ceph.com visitors map here, people all around the world search for it. You can also see some metrics about the Ceph project at their metrics site here.

I’m going to go over the installation of a two node Ceph storage cluster.
You can then build on those storage nodes and choose multiple storage types you want to use.

I personally recommend to use at least two nodes even for a test cluster!


I assume that you have basic linux knowledge and at least two servers with CentOS7 up and running that can reach each other.

In my case my two example servers are having 2vCores and 4GB RAM. The OS used is CentOS7 (to be exact CentOS Linux release 7.2.1511 (Core)).
My examples servers for this tutorial have the following addresses:

Please note the IPs used here are according to RFC5737

You have to decide from what server you want to deploy your cluster.
I use the first server ceph-tutorial-node1 as the “deployment” server.

Preparing your servers for Ceph

Let’s begin first by adding the Epel and Ceph repository to all our used servers.
On all the servers we have, we need to add the Epel package repositories to the repository list.
The following commands add the Epel package repositories and the Epel repository verification key:

sudo yum install -y yum-utils && sudo yum-config-manager --add-repo https://dl.fedoraproject.org/pub/epel/7/x86_64/
sudo yum install --nogpgcheck -y epel-release
sudo rpm --import /etc/pki/rpm-gpg/RPM-GPG-KEY-EPEL-7
sudo rm /etc/yum.repos.d/dl.fedoraproject.org*

Now that our servers have the Epel repository added, we’re finally going to add the Ceph repository.
Use your favorited editor to create a file /etc/yum.repos.d/ceph.repo and copy the following content into it:

name=Ceph noarch packages

Please note You can replace jewel with the Ceph release version you want to install (current latest release is jewel) and el7 with, depending on what CentOS version number you have, for example for CentOS7 it should be el7 and for CentOS6 it should be el6 (other CentOS versions don’t have official rpm builds available and will not work!).

Now we have the Ceph repository in our repository list on all of our servers.
An important step is to configure NTP service on all servers to keep the time synchronized across the storage cluster, but I’m not going to cover this.
Other important things to make sure, before continuing with the tutorial:

Preparing the Ceph deployment

Switch to your deployment server and install the Ceph deployment package ceph-deploy on the deployment server, in my case server ceph-tutorial-node1.
The first commands will update your package database and system. The second command will install the ceph-deploy package on your CentOS server:

sudo yum clean all && sudo yum update
sudo yum install ceph-deploy

Still on the deployment server (in my case ceph-tutorial-node1) create a folder where files for the deployment will be stored. For example:

mkdir ~/my-ceph-cluster
cd ~/my-ceph-cluster

Deploying Ceph storage nodes

To generate the basic Ceph cluster configuration I run for my two servers:

ceph-deploy new ceph-tutorial-node1 ceph-tutorial-node2

Please note Where ceph-tutorial-node1 is you deployment server.

Example output of the ceph-deploy new ceph-tutorial-node1 command:

# ceph-deploy new ceph-tutorial-node1
[ceph_deploy.conf][DEBUG ] found configuration file at: /root/.cephdeploy.conf
[ceph_deploy.cli][INFO ] Invoked (1.5.34): /usr/bin/ceph-deploy new ceph-tutorial-node1
[ceph_deploy.cli][INFO ] ceph-deploy options:
[ceph_deploy.cli][INFO ] username : None
[ceph_deploy.cli][INFO ] func : <function new at 0x1cfb6e0>
[ceph_deploy.cli][INFO ] verbose : False
[ceph_deploy.cli][INFO ] overwrite_conf : False
[ceph_deploy.cli][INFO ] quiet : False
[ceph_deploy.cli][INFO ] cd_conf : <ceph_deploy.conf.cephdeploy.Conf instance at 0x1d56710>
[ceph_deploy.cli][INFO ] cluster : ceph
[ceph_deploy.cli][INFO ] ssh_copykey : True
[ceph_deploy.cli][INFO ] mon : ['ceph-tutorial-node1']
[ceph_deploy.cli][INFO ] public_network : None
[ceph_deploy.cli][INFO ] ceph_conf : None
[ceph_deploy.cli][INFO ] cluster_network : None
[ceph_deploy.cli][INFO ] default_release : False
[ceph_deploy.cli][INFO ] fsid : None
[ceph_deploy.new][DEBUG ] Creating new cluster named ceph
[ceph_deploy.new][INFO ] making sure passwordless SSH succeeds
[ceph-tutorial-node1][DEBUG ] connected to host: ceph-tutorial-node1
[ceph-tutorial-node1][DEBUG ] detect platform information from remote host
[ceph-tutorial-node1][DEBUG ] detect machine type
[ceph-tutorial-node1][DEBUG ] find the location of an executable
[ceph-tutorial-node1][INFO ] Running command: /usr/sbin/ip link show
[ceph-tutorial-node1][INFO ] Running command: /usr/sbin/ip addr show
[ceph-tutorial-node1][DEBUG ] IP addresses found: ['']
[ceph_deploy.new][DEBUG ] Resolving host ceph-tutorial-node1
[ceph_deploy.new][DEBUG ] Monitor ceph-tutorial-node1 at
[ceph_deploy.new][DEBUG ] Monitor initial members are ['ceph-tutorial-node1']
[ceph_deploy.new][DEBUG ] Monitor addrs are ['']
[ceph_deploy.new][DEBUG ] Creating a random mon key...
[ceph_deploy.new][DEBUG ] Writing monitor keyring to ceph.mon.keyring...
[ceph_deploy.new][DEBUG ] Writing initial config to ceph.conf...

Running ls, we see the generated files in the directory:

# ls
ceph.conf ceph-deploy-ceph.log ceph.mon.keyring

Now we need to add an option to the bottom of the default Ceph configuration file ceph.conf.
The options is:

osd pool default size = 2

The option sets the number of replications, for two nodes you have tp set it to 2 else the Ceph cluster will not become healthy.
You can also add other options here, but as this is just a basic tutorial on how to create a Ceph cluster I’m not covering other options than the default.

Now that our Ceph configuration ceph.conf is ready we are going to install Ceph on all servers and create our initial cluster monitor.
The command for deploying Ceph onto my two servers and creating the monitor server on your server chosen above, in my case ceph-tutorial-node1 the commands look like this:

ceph-deploy install --no-adjust-repos ceph-tutorial-node1 ceph-tutorial-node2
ceph-deploy mon create-initial

Errors If the command fails, you need to apply a bug fix with the command sed -i '78s/allow \*/allow/g' /usr/lib/python2.7/site-packages/ceph_deploy/gatherkeys.py. The bug can be seen here).
Please note If the last command exits with errors, check the connectivity of your servers and firewall between the servers (if a firewall is in place between the servers, make sure it allows the Ceph ports, ports: TCP Ceph mon(itor) 6789, Ceph OSDs 6800:7300 (for more information see Ceph Preflight Docs)).

Creating the storage nodes

Use ssh to connect to your servers that you want to be storage nodes and create a directory for the storaged data.
In my case:

ssh ceph-tutorial-node1
sudo mkdir /var/local/osd
ssh ceph-tutorial-node2
sudo mkdir /var/local/osd

Back on your deployment server (in my case ceph-tutorial-node1) we are going to prepare the storage nodes with the following command:

ceph-deploy osd prepare {ceph-node}:/path/to/directory

For my two servers I run:

ceph-deploy osd prepare ceph-tutorial-node1:/var/local/osd ceph-tutorial-node2:/var/local/osd

The ceph-deploy osd prepare prepares the storage location and sets up the Ceph OSD in the background.
The only thing we have to do now is to activate the storage locations (Ceph OSDs) using the activate command:

ceph-deploy osd activate {ceph-node}:/path/to/directory

For my two servers I run:

ceph-deploy osd activate ceph-tutorial-node1:/var/local/osd ceph-tutorial-node2:/var/local/osd

Finalizing your Ceph cluster

Copy the Ceph admin key to the servers you want to be able to configure your Ceph cluster from with the below command:

ceph-deploy admin {deployment-node}

Please note Where {deployment-node} is your server you ran all the commands on.

To check if you Ceph cluster is healthy use the health command:

# ceph health

Please note If it shows something else than HEALTH_OK your cluster is unhealthy and may not even work properly.

That’s all. Now you have a working Ceph storage cluster.


Log file location

In case of other problems, all Ceph log files are located under /var/log/ceph/.

# ls -hl /var/log/ceph/
total 340K
-rw------- 1 root root 126K Aug 8 13:24 ceph.log
-rw-r--r-- 1 root root 196K Aug 8 13:25 ceph-mon.ceph-tutorial-node1.log
-rw-r--r-- 1 root root 5.2K Aug 8 09:26 ceph-osd.0.log

“Clock skew” error

If the health command returns a message like clock skew detected, then your clock on the server(s) is not synced. Please install and configure an NTP server and client(s).