Dell PowerFlex Storage driver

Overview

Dell PowerFlex (formerly named Dell 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 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:

  1. Go to the PowerFlex product documentation page.

  2. On the page, search for the relevant PowerFlex version.

Supported PowerFlex or VxFlex OS Versions

The Dell PowerFlex Block Storage driver has been tested against the following versions of VxFlex OS and PowerFlex and found to be compatible:

  • PowerFlex 3.6.0

  • PowerFlex 4.0.x

  • PowerFlex 4.5.x

Please consult the Official PowerFlex documentation to determine supported operating systems for each version of PowerFlex or VxFlex OS.

Deployment prerequisites

  • 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.

Note

Ubuntu users must follow the specific instructions in the PowerFlex OS Deployment Guide for Ubuntu environments. See the Deploying on Ubuntu Servers section in PowerFlex Deployment Guide. See Official PowerFlex documentation.

Supported operations

  • 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.

Edit the cinder.conf file by adding the configuration below under a new section (for example, [powerflex]) and change the enable_backends setting (in the [DEFAULT] section) to include this new back end. The configuration file is usually located at /etc/cinder/cinder.conf.

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 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>

Oversubscription

Configure the oversubscription ratio by adding the following parameter under the separate section for PowerFlex:

powerflex_max_over_subscription_ratio = <OVER_SUBSCRIPTION_RATIO>

Note

The default value for powerflex_max_over_subscription_ratio is 10.0.

Oversubscription is calculated correctly by the Block Storage service only if the extra specification provisioning:type 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 san_thin_provision option in the configuration file. The default provisioning type will be thin if the option is not specified in the configuration file. To set the default provisioning type thick, set the san_thin_provision option to false in the configuration file, as follows:

san_thin_provision = false

The configuration file is usually located in /etc/cinder/cinder.conf. For a configuration example, see: cinder.conf.

Configuration example

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

Connector configuration

Before using attach/detach volume operations PowerFlex connector must be properly configured. On each node where PowerFlex SDC is installed do the following:

  1. Create /opt/emc/scaleio/openstack/connector.conf if it does not exist.

    $ mkdir -p /opt/emc/scaleio/openstack
    $ touch /opt/emc/scaleio/openstack/connector.conf
    
  2. For each PowerFlex section in the cinder.conf create the same section in the /opt/emc/scaleio/openstack/connector.conf and 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
    

Configuration options

The PowerFlex driver supports these configuration options:

Description of PowerFlex configuration options

Configuration option = Default value

Description

powerflex_allow_migration_during_rebuild = False

(Boolean) Allow volume migration during rebuild.

powerflex_allow_non_padded_volumes = False

(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.

powerflex_max_over_subscription_ratio = 10.0

(Float) max_over_subscription_ratio setting for the driver. Maximum value allowed is 10.0.

powerflex_rest_server_port = 443

(Port(min=0, max=65535)) Gateway REST server port.

powerflex_round_volume_capacity = True

(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

powerflex_server_api_version = None

(String) PowerFlex/ScaleIO API version. This value should be left as the default value unless otherwise instructed by technical support.

powerflex_storage_pools = None

(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

powerflex_unmap_volume_before_deletion = False

(Boolean) Unmap volumes before deletion.

vxflexos_allow_migration_during_rebuild = False

(Boolean) renamed to powerflex_allow_migration_during_rebuild. DEPRECATED

vxflexos_allow_non_padded_volumes = False

(Boolean) renamed to powerflex_allow_non_padded_volumes. DEPRECATED

vxflexos_max_over_subscription_ratio = 10.0

(Float) renamed to powerflex_max_over_subscription_ratio. DEPRECATED

vxflexos_rest_server_port = 443

(Port(min=0, max=65535)) renamed to powerflex_rest_server_port. DEPRECATED

vxflexos_round_volume_capacity = True

(Boolean) renamed to powerflex_round_volume_capacity. DEPRECATED

vxflexos_server_api_version = None

(String) renamed to powerflex_server_api_version. DEPRECATED

vxflexos_storage_pools = None

(String) renamed to powerflex_storage_pools. DEPRECATED

vxflexos_unmap_volume_before_deletion = False

(Boolean) renamed to powerflex_round_volume_capacity. DEPRECATED

Volume Types

Volume types can be used to specify characteristics of volumes allocated via the PowerFlex Driver. These characteristics are defined as Extra Specs within Volume Types.

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 pool_name Extra Spec to the volume type extra-specs and setting the value to the requested protection_domain:storage_pool.

$ 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:

maxIOPS

The QoS I/O rate limit. If not set, the I/O rate will be unlimited. The setting must be larger than 10.

maxIOPSperGB

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.

maxBWS

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.

maxBWSperGB

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 maxIOPSperGB or maxBWSperGB.

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.

provisioning:type

storage pool supports compression

yes (PowerFlex 3.0 FG pool)

no (other pools)

compressed

thin with compression

thin

thin

thin

thin

thick

thin

thick

not set

thin

thin

Note

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.

Prerequisites

  • 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.

Configure replication

  1. Enable replication in cinder.conf file.

    To enable replication feature for storage backend replication_device must 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 - powerflex_rest_server_port.

      • SSL certificate verification - driver_ssl_cert_verify and driver_ssl_cert_path.

    For more information see Configuration options.

  2. 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
    
  3. Set PowerFlex Replication Consistency Group name for volume type.

    $ openstack volume type set --property powerflex:replication_cg=<replication_cg name> \
        powerflex_replicated
    
  4. 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.conf you should set pool_name property for volume type with appropriate Protection Domain and Storage Pool. See PowerFlex Protection Domain and Storage Pool.

Failover host

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 --backend_id default:

$ 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.

Known limitations

  • 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.

Migrate volume

Volume migration is performed by issuing the following command:

$ cinder migrate <volume> <host>

Note

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

  1. Create a file with below contents:

    parameter_defaults:
      NovaComputeOptVolumes:
        - /opt/emc/scaleio:/opt/emc/scaleio
      CinderVolumeOptVolumes:
        - /opt/emc/scaleio:/opt/emc/scaleio
      GlanceApiOptVolumes:
        - /opt/emc/scaleio:/opt/emc/scaleio
    

    Name it whatever you like, e.g. powerflex_volumes.yml.

  2. Use -e to include this customization file to deploy command.

  3. Install the Storage Data Client (SDC) on all nodes after deploying the overcloud.