StorPool Integration with OpenStack¶

This document describes the ways to configure the OpenStack Cinder block storage component to use the StorPool distributed storage as a backend and provide StorPool-backed volumes to the other OpenStack services such as Nova Compute.

StorPool Distributed Storage¶

Introduction¶

StorPool is distributed data storage software running on standard x86 servers. StorPool aggregates the performance and capacity of all drives into a shared pool of storage distributed among the servers. Within this storage pool the user creates thin-provisioned volumes that are exposed to the clients as block devices. StorPool consists of two parts wrapped in one package - a server and a client. The StorPool server allows a hypervisor to act as a storage node, while the StorPool client allows a hypervisor node to access the storage pool and act as a compute node. In OpenStack terms the StorPool solution allows each hypervisor node to be both a storage and a compute node simultaneously.

Concepts¶

In a StorPool cluster the user data is stored in virtual volumes managed in a transparent way by the StorPool servers. When needed, a volume is attached to a client host using the StorPool client device driver and exposed to processes running on the host as a block device.

Components¶

Hosts that will use the StorPool volumes only need to run the storpool_beacon (StorPool cluster presence and quorum) and the storpool_block (StorPool block device client) services.

On a hyperconverged setup, the hosts will also run the storpool_server (StorPool data storage) and storpool_mgmt (StorPool management API) services. On setups with differentiated storage and client hosts, these services will only need to run on the storage cluster.

Configuring the OpenStack Cinder StorPool volume backend¶

An OpenStack environment that uses StorPool as a storage backend only needs the client services; the actual disks that store the data may be on a separate storage cluster.

This section describes the configuration of the OpenStack Cinder (block storage) service. Note that for OpenStack releases before Queens (February 2018), the StorPool Cinder drivers may need to be installed beforehand; see the “Installing the StorPool drivers” section for details.

Configuring the StorPool Cinder backend¶

Once the StorPool Cinder volume driver is installed (as part of OpenStack or, for earlier releases, separately), a driver backend section must be created in the /etc/cinder/cinder.conf file:

[storpool]
volume_driver=cinder.volume.drivers.storpool.StorPoolDriver
storpool_template=hybrid-r3
report_discard_supported=True

If StorPool should be the only storage backend, all the enabled_backends and volume_backend lines should be removed from the cinder.conf file’s [DEFAULT] section and a single line should be added:

enabled_backends=storpool

If other volume backends should coexist with StorPool, make sure that storpool is listed in an enabled_backends setting.

Then restart the relevant Cinder services:

# service cinder-volume restart
# service cinder-scheduler restart

Using StorPool templates via Cinder volume types¶

If a StorPool cluster has several different templates defined (for reasons ranging from quality of service to fault sets and availability zones), they may be exposed to Cinder through the use of volume types. For each StorPool template, define a volume type and specify the StorPool template name as its storpool_template property:

# openstack volume type create --property volume_backend_name=storpool --property storpool_template='hybrid-r2' hybrid-r2
# openstack volume type create --property volume_backend_name=storpool --property storpool_template='hdd-only' hdd-only

The names of the available StorPool templates may be viewed with the storpool template list command if the StorPool CLI is installed.

Creating volumes and snapshots¶

To create a new, blank volume in one of the previously defined volume types:

# openstack volume create --type hybrid --size 10 test-volume

To make sure that there is actually a StorPool volume created, first obtain the OpenStack internal volume identifier:

# openstack volume show -f value -c id test-volume
cc2b3692-a2b6-4ad7-82ef-6a37d3e0f4da

# storpool volume list | fgrep -e 'cc2b3692-a2b6-4ad7-82ef-6a37d3e0f4da'
| os--volume-cc2b3692-a2b6-4ad7-82ef-6a37d3e0f4da |   10 GB |     3 | hdd        | hdd        | ssd        |

To create a volume from an existing image stored in Glance:

# openstack volume create --type hybrid --size 1 --image cirros-0.4.0 cirros-v

To create a snapshot of this volume:

# openstack volume snapshot create --name cirros cirros-v

Configuring the OpenStack Nova StorPool volume attachment driver¶

This section describes the configuration of the OpenStack Nova (compute) service to use StorPool-backed volumes created by Cinder. Note that for OpenStack releases before Queens (February 2018), the StorPool Nova drivers may need to be installed beforehand; see the “Installing the StorPool drivers” section for details.

Configuring the Nova volume attachment driver¶

Edit the /etc/nova/nova.conf file. If there is a volume_drivers line in the [DEFAULT] section, add to it the following definition. If there is no such line, add it:

volume_drivers=storpool=nova.virt.libvirt.volume.LibvirtStorPoolVolumeDriver

Then restart the relevant Nova services:

# service nova-compute restart
# service nova-api-metadata restart

Using StorPool volumes as instance root disks¶

The preferred way to use StorPool volumes as root disks for Nova instances is to create a volume-backed instance and a new volume for it to use.

Choose a Glance image to use as the root disk.
Create a Cinder volume from that Glance image.
Create a snapshot of that Cinder volume.
Create a Nova instance and tell it to create a new volume from the snapshot.

Step 4 may be repeated for many instances; a new volume shall be created for each instance’s root disk, but within the StorPool cluster all these volumes will be descended from the one holding the Cinder snapshot, so there shall be no needless copying of data and no overuse of disk space.

Installing the StorPool drivers¶

For the OpenStack releases before Queens, namely Mitaka (April 2016), Newton (October 2016), Ocata (February 2017), and Pike (August 2017), the StorPool drivers for Cinder and Nova are not part of the OpenStack distribution and must be installed separately. The tools referenced in this section may be obtained from the StorPool OpenStack Integration Git repository.

Preliminary setup¶

Set up an OpenStack cluster.
Set up a StorPool cluster.
Set up at least the StorPool client (the storpool_beacon and storpool_block services) on the OpenStack controller node and on each of the hypervisor nodes.
On the controller and all the hypervisor nodes, install the ``storpool` <https://github.com/storpool/python-storpool>`_ and ``storpool-spopenstack` <https://github.com/storpool/python-storpool-spopenstack>`_ Python packages:
```
# Install pip (the Python package installer) first
yum install python-pip
# ...or...
apt-get install --no-install-recommends python-pip

pip install storpool
pip install storpool.spopenstack
```
On the controller and all the hypervisor nodes, clone the StorPool OpenStack Integration Git repository or copy it from some other host where it has been cloned:
```
# Clone the StorPool OpenStack Integration repository
git clone https://github.com/storpool/storpool-openstack-integration.git
```

Set up the Cinder volume backend¶

Change into the storpool-openstack-integration/ directory prepared in the last step of the “Preliminary setup” section above.
Let the StorPool OpenStack integration suite find your Cinder drivers and apply the StorPool integration changes: and make the necessary modifications in its own work directory:
```
./sp-openstack check
./sp-openstack install cinder os_brick
```
Proceed as per the “Configuring the OpenStack Cinder StorPool volume backend” section above.

Set up the Nova volume attachment driver (on each hypevisor node)¶

Change into the storpool-openstack-integration/ directory prepared in the last step of the “Preliminary setup” section above.
Make sure that the Python modules ``storpool` <https://github.com/storpool/python-storpool>`_ and ``storpool.spopenstack` <https://github.com/storpool/python-storpool-spopenstack>`_ have also been installed on all the hypervisor nodes (see the “Preliminary setup” section above).
Let the StorPool OpenStack integration suite find your Nova drivers and and make the necessary modifications in its own work directory:
```
./sp-openstack check
./sp-openstack install nova
```
Proceed as per the “Configuring the OpenStack Nova StorPool volume attachment driver” section above.