Single-node install

This tutorial will show how to install MicroCeph on a single machine, thereby creating a single-node “cluster”.

The above will be achieved through the use of loop files placed on the root disk, which is a convenient way for setting up small test and development clusters.

Warning

Using dedicated block devices will result in the best IOPS performance for connecting clients. Basing a Ceph cluster on a single disk also necessarily leads to a common failure domain for all OSDs. For these reasons, loop files should not be used in production environments.

Install the software

Install the most recent stable release of MicroCeph:

sudo snap install microceph

Next, prevent the software from being auto-updated:

sudo snap refresh --hold microceph

Caution

Allowing the snap to be auto-updated can lead to unintended consequences. In enterprise environments especially, it is better to research the ramifications of software changes before those changes are implemented.

Initialise the cluster

Begin by initialising the cluster with the cluster bootstrap command:

sudo microceph cluster bootstrap

Then look at the status of the cluster with the status command:

sudo microceph status

It should look similar to the following:

MicroCeph deployment summary:
- node-mees (10.246.114.49)
    Services: mds, mgr, mon
      Disks: 0

Here, the machine’s hostname of ‘node-mees’ is given along with its IP address of ‘10.246.114.49’. The MDS, MGR, and MON services are running but there is not yet any storage available.

Add storage

Three OSDs will be required to form a minimal Ceph cluster. In a production system, typically we would assign a physical block device to an OSD. However for this tutorial, we will make use of file backed OSDs for simplicity.

Add the three file-backed OSDs to the cluster by using the disk add command. In the example, three 4GiB files are being created:

sudo microceph disk add loop,4G,3

Note

Although you can adjust the file size and file number to your needs, with a recommended minimum of 2GiB per OSD, there is no obvious benefit to running more than three OSDs via loop files. Be wary that an OSD, whether based on a physical device or a file, is resource intensive.

Recheck status:

sudo microceph status

The output should now show three disks and the additional presence of the OSD service:

MicroCeph deployment summary:
- node-mees (10.246.114.49)
    Services: mds, mgr, mon, osd
      Disks: 3

Manage the cluster

Your Ceph cluster is now deployed and can be managed by following the resources found in the How-to section.

The cluster can also be managed using native Ceph tooling if snap-level commands are not yet available for a desired task:

sudo ceph status

The cluster built during this tutorial gives the following output:

cluster:
  id:     4c2190cd-9a31-4949-a3e6-8d8f60408278
  health: HEALTH_OK

services:
  mon: 1 daemons, quorum node-mees (age 7d)
  mgr: node-mees(active, since 7d)
  osd: 3 osds: 3 up (since 7d), 3 in (since 7d)

data:
  pools:   1 pools, 1 pgs
  objects: 2 objects, 577 KiB
  usage:   96 MiB used, 2.7 TiB / 2.7 TiB avail
  pgs:     1 active+clean