Dell EMC PowerFlex Storage driver¶
Dell EMC PowerFlex (formerly named Dell EMC ScaleIO/VxFlex OS) is a software-only solution that uses existing servers local disks and LAN to create a virtual SAN that has all of the benefits of external storage, but at a fraction of the cost and complexity. Using the driver, Block Storage hosts can connect to a PowerFlex Storage cluster.
The Dell EMC PowerFlex Cinder driver is designed and tested to work with both PowerFlex and with ScaleIO. The configuration options are identical for both PowerFlex and ScaleIO.
Official PowerFlex documentation¶
To find the PowerFlex documentation:
Go to the PowerFlex product documentation page.
From the left-side panel, select the relevant PowerFlex version.
Supported PowerFlex, VxFlex OS and ScaleIO Versions¶
The Dell EMC PowerFlex Block Storage driver has been tested against the following versions of ScaleIO, VxFlex OS and PowerFlex and found to be compatible:
VxFlex OS 2.6.x
VxFlex OS 3.0.x
Please consult the Official PowerFlex documentation to determine supported operating systems for each version of PowerFlex, VxFlex OS or ScaleIO.
The PowerFlex Gateway must be installed and accessible in the network. For installation steps, refer to the Preparing the installation Manager and the Gateway section in PowerFlex Deployment Guide. See Official PowerFlex documentation.
PowerFlex Storage Data Client (SDC) must be installed on all OpenStack nodes.
Ubuntu users must follow the specific instructions in the PowerFlex
OS Deployment Guide for Ubuntu environments. See the
on Ubuntu Servers section in PowerFlex Deployment Guide. See
Official PowerFlex documentation.
Create, delete, clone, attach, detach, migrate, manage, and unmanage volumes
Create, delete, manage, and unmanage volume snapshots
Create a volume from a snapshot
Revert a volume to a snapshot
Copy an image to a volume
Copy a volume to an image
Extend a volume
Get volume statistics
Create, list, update, and delete consistency groups
Create, list, update, and delete consistency group snapshots
OpenStack replication v2.1 support
PowerFlex Block Storage driver configuration¶
This section explains how to configure and connect the block storage nodes to a PowerFlex storage cluster.
cinder.conf file by adding the configuration below under
a new section (for example,
[powerflex]) and change the
setting (in the
[DEFAULT] section) to include this new back end.
The configuration file is usually located at
For a configuration example, refer to the example cinder.conf.
PowerFlex driver name¶
Configure the driver name by adding the following parameter:
volume_driver = cinder.volume.drivers.dell_emc.powerflex.driver.PowerFlexDriver
PowerFlex Gateway server IP¶
The PowerFlex Gateway provides a REST interface to PowerFlex.
Configure the Gateway server IP address by adding the following parameter:
san_ip = <PowerFlex GATEWAY IP>
PowerFlex Storage Pools¶
Multiple Storage Pools and Protection Domains can be listed for use by the virtual machines. The list should include every Protection Domain and Storage Pool pair that you would like Cinder to utilize.
To retrieve the available Storage Pools, use the command scli --query_all and search for available Storage Pools.
Configure the available Storage Pools by adding the following parameter:
powerflex_storage_pools = <Comma-separated list of protection domain:storage pool name>
PowerFlex user credentials¶
Block Storage requires a PowerFlex user with administrative privileges. Dell EMC recommends creating a dedicated OpenStack user account that has an administrative user role.
Refer to the PowerFlex User Guide for details on user account management.
Configure the user credentials by adding the following parameters:
san_login = <POWERFLEX_USER> san_password = <POWERFLEX_PASSWD>
Configure the oversubscription ratio by adding the following parameter under the separate section for PowerFlex:
powerflex_max_over_subscription_ratio = <OVER_SUBSCRIPTION_RATIO>
The default value for
Oversubscription is calculated correctly by the Block Storage service
only if the extra specification
appears in the volume type regardless of the default provisioning type.
Maximum oversubscription value supported for PowerFlex is 10.0.
Default provisioning type¶
If provisioning type settings are not specified in the volume type,
the default value is set according to the
option in the configuration file. The default provisioning type
thin if the option is not specified in the configuration
file. To set the default provisioning type
san_thin_provision option to
in the configuration file, as follows:
san_thin_provision = false
The configuration file is usually located in
For a configuration example, see:
cinder.conf example file
You can update the
cinder.conf file by editing the necessary
parameters as follows:
[DEFAULT] enabled_backends = powerflex [powerflex] volume_driver = cinder.volume.drivers.dell_emc.powerflex.driver.PowerFlexDriver volume_backend_name = powerflex san_ip = GATEWAY_IP powerflex_storage_pools = Domain1:Pool1,Domain2:Pool2 san_login = POWERFLEX_USER san_password = POWERFLEX_PASSWD san_thin_provision = false
Before using attach/detach volume operations PowerFlex connector must be properly configured. On each node where PowerFlex SDC is installed do the following:
/opt/dellemc/powerflex/openstack/connector.confif it does not exist.
$ mkdir -p /opt/dellemc/powerflex/openstack $ touch /opt/dellemc/powerflex/openstack/connector.conf
For each PowerFlex section in the
cinder.confcreate the same section in the
/opt/dellemc/powerflex/openstack/connector.confand populate it with passwords. Example:
[powerflex] san_password = POWERFLEX_PASSWD replicating_san_password = REPLICATION_SYSTEM_POWERFLEX_PASSWD # if applicable [powerflex-new] san_password = SIO2_PASSWD replicating_san_password = REPLICATION_SYSTEM_SIO2_PASSWD # if applicable
The PowerFlex driver supports these configuration options:
Configuration option = Default value
(Boolean) Allow volume migration during rebuild.
(Boolean) Allow volumes to be created in Storage Pools when zero padding is disabled. This option should not be enabled if multiple tenants will utilize volumes from a shared Storage Pool.
(Float) max_over_subscription_ratio setting for the driver. Maximum value allowed is 10.0.
(Port(min=0, max=65535)) Gateway REST server port.
(Boolean) Round volume sizes up to 8GB boundaries. PowerFlex/VxFlex OS requires volumes to be sized in multiples of 8GB. If set to False, volume creation will fail for volumes not sized properly
(String) PowerFlex/ScaleIO API version. This value should be left as the default value unless otherwise instructed by technical support.
(String) Storage Pools. Comma separated list of storage pools used to provide volumes. Each pool should be specified as a protection_domain_name:storage_pool_name value
(Boolean) Unmap volumes before deletion.
(Boolean) renamed to powerflex_allow_migration_during_rebuild. DEPRECATED
(Boolean) renamed to powerflex_allow_non_padded_volumes. DEPRECATED
(Float) renamed to powerflex_max_over_subscription_ratio. DEPRECATED
(Port(min=0, max=65535)) renamed to powerflex_rest_server_port. DEPRECATED
(Boolean) renamed to powerflex_round_volume_capacity. DEPRECATED
(String) renamed to powerflex_server_api_version. DEPRECATED
(String) renamed to powerflex_storage_pools. DEPRECATED
(Boolean) renamed to powerflex_round_volume_capacity. DEPRECATED
Volume types can be used to specify characteristics of volumes allocated via
the PowerFlex Driver. These characteristics are defined as
PowerFlex Protection Domain and Storage Pool¶
When multiple storage pools are specified in the Cinder configuration,
users can specify which pool should be utilized by adding the
Extra Spec to the volume type extra-specs and setting the value to the
$ openstack volume type create powerflex_type_1 $ openstack volume type set --property volume_backend_name=powerflex powerflex_type_1 $ openstack volume type set --property pool_name=Domain2:Pool2 powerflex_type_1
PowerFlex thin provisioning support¶
The Block Storage driver supports creation of thin-provisioned and thick-provisioned volumes. The provisioning type settings can be added as an extra specification of the volume type, as follows:
$ openstack volume type create powerflex_type_thick $ openstack volume type set --property provisioning:type=thick powerflex_type_thick
PowerFlex QoS support¶
QoS support for the PowerFlex driver includes the ability to set the following capabilities:
The QoS I/O rate limit. If not set, the I/O rate will be unlimited. The setting must be larger than 10.
The QoS I/O rate limit. The limit will be calculated by the specified value multiplied by the volume size. The setting must be larger than 10.
The QoS I/O bandwidth rate limit in KBs. If not set, the I/O bandwidth rate will be unlimited. The setting must be a multiple of 1024.
The QoS I/O bandwidth rate limit in KBs. The limit will be calculated by the specified value multiplied by the volume size. The setting must be a multiple of 1024.
The QoS keys above must be created and associated with a volume type. For example:
$ openstack volume qos create qos-limit-iops --consumer back-end --property maxIOPS=5000 $ openstack volume type create powerflex_limit_iops $ openstack volume qos associate qos-limit-iops powerflex_limit_iops
The driver always chooses the minimum between the QoS keys value
and the relevant calculated value of
Since the limits are per SDC, they will be applied after the volume is attached to an instance, and thus to a compute node/SDC.
PowerFlex compression support¶
Starting from version 3.0, PowerFlex supports volume compression. By default driver will create volumes without compression. In order to create a compressed volume, a volume type which enables compression support needs to be created first:
$ openstack volume type create powerflex_compressed $ openstack volume type set --property provisioning:type=compressed powerflex_compressed
If a volume with this type is scheduled to a storage pool which doesn’t
support compression, then
thin provisioning will be used.
See table below for details.
storage pool supports compression
yes (PowerFlex 3.0 FG pool)
no (other pools)
thin with compression
PowerFlex 3.0 Fine Granularity storage pools don’t support thick provisioned volumes.
You can add property
compression_support='<is> True' to volume type to
limit volumes allocation only to data pools which supports compression.
$ openstack volume type set --property compression_support='<is> True' powerflex_compressed
PowerFlex replication support¶
Starting from version 3.5, PowerFlex supports volume replication.
PowerFlex replication components must be installed on source and destination systems.
Source and destination systems must have the same configuration for Protection Domains and their Storage Pools (i.e. names, zero padding, etc.).
Source and destination systems must be paired and have at least one Replication Consistency Group created.
See Official PowerFlex documentation for instructions.
Enable replication in
To enable replication feature for storage backend
replication_devicemust be set as below:
[DEFAULT] enabled_backends = powerflex [powerflex] volume_driver = cinder.volume.drivers.dell_emc.powerflex.driver.PowerFlexDriver volume_backend_name = powerflex san_ip = GATEWAY_IP powerflex_storage_pools = Domain1:Pool1,Domain2:Pool2 san_login = POWERFLEX_USER san_password = POWERFLEX_PASSWD san_thin_provision = false replication_device = backend_id:powerflex_repl, san_ip: REPLICATION_SYSTEM_GATEWAY_IP, san_login: REPLICATION_SYSTEM_POWERFLEX_USER, san_password: REPLICATION_SYSTEM_POWERFLEX_PASSWD
Only one replication device is supported for storage backend.
The following parameters are optional for replication device:
REST API port -
SSL certificate verification -
For more information see Configuration options.
Create volume type for volumes with replication enabled.
$ openstack volume type create powerflex_replicated $ openstack volume type set --property replication_enabled='<is> True' powerflex_replicated
Set PowerFlex Replication Consistency Group name for volume type.
$ openstack volume type set --property powerflex:replication_cg=<replication_cg name> \ powerflex_replicated
Set Protection Domain and Storage Pool if multiple Protection Domains are specified.
PowerFlex Replication Consistency Group is created between source and destination Protection Domains. If more than one Protection Domain is specified in
cinder.confyou should set
pool_nameproperty for volume type with appropriate Protection Domain and Storage Pool. See PowerFlex Protection Domain and Storage Pool.
In the event of a disaster, or where there is a required downtime the administrator can issue the failover host command:
$ cinder failover-host cinder_host@powerflex --backend_id powerflex_repl
After issuing Cinder failover-host command Cinder will switch to configured replication device, however to get existing instances to use this target and new paths to volumes it is necessary to first shelve Nova instances and then unshelve them, this will effectively restart the Nova instance and re-establish data paths between Nova instances and the volumes.
$ nova shelve <server> $ nova unshelve [--availability-zone <availability_zone>] <server>
If the primary system becomes available, the administrator can initiate
failback operation using
$ cinder failover-host cinder_host@powerflex --backend_id default
PowerFlex storage-assisted volume migration¶
Starting from version 3.0, PowerFlex supports storage-assisted volume migration.
Migration between different backends is not supported.
For migration from Medium Granularity (MG) to Fine Granularity (FG) storage pool zero padding must be enabled on the MG pool.
For migration from MG to MG pool zero padding must be either enabled or disabled on both pools.
In the above cases host-assisted migration will be perfomed.
Volume migration is performed by issuing the following command:
$ cinder migrate <volume> <host>
Volume migration has a timeout of 3600 seconds (1 hour).
It is done to prevent from endless waiting for migration to
complete if something unexpected happened. If volume still is in
migration after timeout has expired, volume status will be changed to
maintenance to prevent future operations with this volume. The
corresponding warning will be logged.
In this situation the status of the volume should be checked on the storage side. If volume migration succeeded, its status can be changed manually:
$ cinder reset-state --state available <volume>
Using PowerFlex Storage with a containerized overcloud¶
Create a file with below contents:
parameter_defaults: NovaComputeOptVolumes: - /opt/dellemc/powerflex:/opt/dellemc/powerflex CinderVolumeOptVolumes: - /opt/dellemc/powerflex:/opt/dellemc/powerflex GlanceApiOptVolumes: - /opt/dellemc/powerflex:/opt/dellemc/powerflex
Name it whatever you like, e.g.
-eto include this customization file to deploy command.
Install the Storage Data Client (SDC) on all nodes after deploying the overcloud.