Published on

Setup a 3 Node Ceph Storage Cluster on Ubuntu 16

Authors

For some time now, I wanted to do a setup of Ceph, and I finally got the time to do it. This setup was done on Ubuntu 16.04

What is Ceph

Ceph is a storage platform that implements object storage on a single distributed computer cluster and provides interfaces for object, block and file-level storage.

  • Object Storage:

Ceph provides seemless access to objects via native language bindings or via the REST interface, RadosGW and also compatible for applications written for S3 and Swift.

  • Block Storage:

Ceph's Rados Block Device (RBD) provides access to block device images that are replicated and striped across the storage cluster.

  • File System:

Ceph provides a network file system (CephFS) that aims for high performance.

Our Setup

We will have 4 nodes. 1 Admin node where we will deploy our cluster with, and 3 nodes that will hold the data:

  • ceph-admin (10.0.8.2)
  • ceph-node1 (10.0.8.3)
  • ceph-node2 (10.0.8.4)
  • ceph-node3 (10.0.8.5)

Host Entries

If you don't have dns for your servers, setup the /etc/hosts file so that the names can resolves to the ip addresses:

10.0.8.2 ceph-admin
10.0.8.3 ceph-node1
10.0.8.4 ceph-node2
10.0.8.5 ceph-node3

User Accounts and Passwordless SSH

Setup the ceph-system user accounts on all the servers:

$ useradd -d /home/ceph-system -s /bin/bash -m ceph-system
$ passwd ceph-system

Setup the created user part of the sudoers that is able to issue sudo commands without a pssword:

$ echo "ceph-system ALL = (root) NOPASSWD:ALL" | sudo tee /etc/sudoers.d/ceph-system
$ chmod 0440 /etc/sudoers.d/ceph-system

Switch user to ceph-system and generate SSH keys and copy the keys from the ceph-admin server to the ceph-nodes:

$ sudo su - ceph-system
$ ssh-keygen -t rsa -f ~/.ssh/id_rsa -P ""
$ ssh-copy-id ceph-system@ceph-node1
$ ssh-copy-id ceph-system@ceph-node2
$ ssh-copy-id ceph-system@ceph-node3
$ ssh-copy-id ceph-system@ceph-admin

Pre-Requisite Software:

Install Python and Ceph Deploy on each node:

$ sudo apt-get install python -y
$ sudo apt install ceph-deploy -y

Note: Please skip this section if you have additional disks on your servers.

The instances that im using to test this setup only has one disk, so I will be creating loop block devices using allocated files. This is not recommended as when the disk fails, all the (files/block device images) will be gone with that. But since im demonstrating this, I will create the block devices from a file:

I will be creating a 12GB file on each node

$ sudo mkdir /raw-disks 
$ sudo dd if=/dev/zero of=/raw-disks/rd0 bs=1M count=12288

The use losetup to create the loop0 block device:

$ sudo losetup /dev/loop0 /raw-disks/rd0

As you can see the loop device is showing when listing the block devices:

$ lsblk
NAME      MAJ:MIN RM  SIZE RO TYPE MOUNTPOINT
loop0       7:0    0   12G  0 loop

Install Ceph

Now let's install ceph using ceph-deploy to all our nodes:

$ sudo apt update && sudo apt upgrade -y
$ ceph-deploy install ceph-admin ceph-node1 ceph-node2 ceph-node3

The version I was running at the time:

$ ceph --version
ceph version 10.2.9

Initialize Ceph

Initialize the Cluster with 3 Monitors:

$ ceph-deploy new ceph-node1 ceph-node2 ceph-node3

Add the initial monitors and gather the keys from the previous command:

$ ceph-deploy mon create-initial

At this point, we should be able to scan the block devices on our nodes:

$ ceph-deploy disk list ceph-node3
[ceph-node3][INFO  ] Running command: sudo /usr/sbin/ceph-disk list
[ceph-node3][DEBUG ] /dev/loop0 other

Prepare the Disks:

First we will zap the block devices and then prepare to create the partitions:

$ ceph-deploy disk zap ceph-node1:/dev/loop0 ceph-node2:/dev/loop0 ceph-node3:/dev/loop0
$ ceph-deploy osd prepare ceph-node1:/dev/loop0 ceph-node2:/dev/loop0 ceph-node3:/dev/loop0
[ceph_deploy.osd][DEBUG ] Host ceph-node1 is now ready for osd use.
[ceph_deploy.osd][DEBUG ] Host ceph-node2 is now ready for osd use.
[ceph_deploy.osd][DEBUG ] Host ceph-node3 is now ready for osd use.

When you scan the nodes for their disks, you will notice that the partitions has been created:

$ ceph-deploy disk list ceph-node1 
[ceph-node1][DEBUG ] /dev/loop0p2 ceph journal, for /dev/loop0p1 
[ceph-node1][DEBUG ] /dev/loop0p1 ceph data, active, cluster ceph, osd.0, journal /dev/loop0p2

Now let's activate the OSD's by using the data partitions:

$ ceph-deploy osd activate ceph-node1:/dev/loop0p1 ceph-node2:/dev/loop0p1 ceph-node3:/dev/loop0p1

Redistribute Keys:

Copy the configuration files and admin key to your admin node and ceph data nodes:

$ ceph-deploy admin ceph-admin ceph-node1 ceph-node2 ceph-node3

If you would like to add more OSD's (not tested):

$ ceph-deploy disk zap ceph-node1:/dev/loop1 ceph-node2:/dev/loop1 ceph-node3:/dev/loop1
$ ceph-deploy osd prepare ceph-node1:/dev/loop1 ceph-node2:/dev/loop1 ceph-node3:/dev/loop1 
$ ceph-deploy osd activate ceph-node2:/dev/loop1p1:/dev/loop1p2 ceph-node2:/dev/loop1p1:/dev/loop1p2 ceph-node3:/dev/loop1p1:/dev/loop1p2
$ ceph-deploy admin ceph-node1 ceph-node2 ceph-node3

Ceph Status:

Have a look at your cluster status:

$ sudo ceph -s
    cluster 8d704c8a-ac19-4454-a89f-89a5d5b7d94d
     health HEALTH_OK
     monmap e1: 3 mons at {ceph-node1=10.0.8.3:6789/0,ceph-node2=10.0.8.4:6789/0,ceph-node3=10.0.8.5:6789/0}
            election epoch 10, quorum 0,1,2 ceph-node2,ceph-node3,ceph-node1
     osdmap e14: 3 osds: 3 up, 3 in
            flags sortbitwise,require_jewel_osds
      pgmap v29: 64 pgs, 1 pools, 0 bytes data, 0 objects
            100 MB used, 18298 MB / 18398 MB avail
                  64 active+clean

Everything looks good. Also change the permissions on this file, on all the nodes in order to execute the ceph, rados commands:

$ sudo chmod +r /etc/ceph/ceph.client.admin.keyring

Storage Pools:

List your pool in your Ceph Cluster:

$ rados lspools
rbd

Let's create a new storage pool called mypool:

$ ceph osd pool create mypool 32 32 
pool 'mypool' created

Let's the list the storage pools again:

$ rados lspools 
rbd 
mypool

You can also use the ceph command to list the pools:

$ ceph osd pool ls 
rbd 
mypool

Create a Block Device Image:

$ rbd create --size 1024 mypool/disk1 --image-feature layering

List the Block Device Images under your Pool:

$ rbd list mypool
disk1

Retrieve information from your image:

$ rbd info mypool/disk1
rbd image 'disk1':
        size 1024 MB in 256 objects
        order 22 (4096 kB objects)
        block_name_prefix: rbd_data.1021643c9869
        format: 2
        features: layering
        flags:
        create_timestamp: Thu Jun  7 23:48:23 2018

Create a local mapping of the image to a block device:

$ sudo rbd map mypool/disk1
/dev/rbd0

Now we have a block device available at /dev/rbd0. Go ahead and mount it to /mnt:

$ sudo mount /dev/rbd0 /mnt

We can then see it when we list our mounted disk partitions:

$ df -h
Filesystem      Size  Used Avail Use% Mounted on
/dev/sda1        19G   13G  5.2G  72% /
/dev/rbd0       976M  1.3M  908M   1% /mnt

We can also resize the disk on the fly, let's resize it from 1GB to 2GB:

$ rbd resize mypool/disk1 --size 2048
Resizing image: 100% complete...done.

To grow the space we can use resize2fs for ext4 partitions and xfs_growfs for xfs partitions:

$ sudo resize2fs /dev/rbd0 
resize2fs 1.42.13 (17-May-2015)
Filesystem at /dev/rbd0 is mounted on /mnt; on-line resizing required
old_desc_blocks = 1, new_desc_blocks = 1
The filesystem on /dev/rbd0 is now 524288 (4k) blocks long.

When we look at our mounted partitions, you will notice that the size of our mounted partition has been increased in size:

$ df -h
Filesystem      Size  Used Avail Use% Mounted on
/dev/sda1        19G   13G  5.2G   72% /
/dev/rbd0       2.0G  1.5M  1.9G   1% /mnt

Object Storage RadosGW

Let's create a new pool where we will store our objects:

$ ceph osd pool create object-pool 32 32
pool 'object-pool' created

We will now create a local file, push the file to our object storage service, then delete our local file, download the file as a file with a different name, and read the contents:

Create the local file:

$ echo "ok" > test.txt

Push the local file to our pool in our object storage:

$ rados put objects/data/test.txt ./test.txt --pool object-pool

List the pool (note that this can be executed from any node):

$ $ rados ls --pool object-pool
objects/data/test.txt

Delete the local file, download the file from our object storage and read the contents:

$ rm -rf test.txt 

$ rados get objects/data/test.txt ./newfile.txt --pool object-pool

$ cat ./newfile.txt 
ok

View the disk space from our storage-pool:

$ rados df --pool object-pool
pool name                 KB      objects       clones     degraded      unfound           rd        rd KB           wr        wr KB
object-pool                1            1            0            0            0            0            0            1            1
  total used          261144           37
  total avail       18579372
  total space       18840516

Resources:

Discuss on Twitter • View on GitHub