SolidFire¶
The SolidFire Cluster is a high performance all SSD iSCSI storage device that provides massive scale out capability and extreme fault tolerance. A key feature of the SolidFire cluster is the ability to set and modify during operation specific QoS levels on a volume for volume basis. The SolidFire cluster offers this along with de-duplication, compression, and an architecture that takes full advantage of SSDs.
To configure the use of a SolidFire cluster with Block Storage, modify your
cinder.conf file as follows:
volume_driver = cinder.volume.drivers.solidfire.SolidFireDriver
san_ip = 172.17.1.182 # the address of your MVIP
san_login = sfadmin # your cluster admin login
san_password = sfpassword # your cluster admin password
sf_account_prefix = '' # prefix for tenant account creation on solidfire cluster
Warning
Older versions of the SolidFire driver (prior to Icehouse) created a unique
account prefixed with $cinder-volume-service-hostname-$tenant-id on the
SolidFire cluster for each tenant. Unfortunately, this account formation
resulted in issues for High Availability (HA) installations and
installations where the cinder-volume service can move to a new node.
The current default implementation does not experience this issue as no
prefix is used. For installations created on a prior release, the OLD
default behavior can be configured by using the keyword hostname in
sf_account_prefix.
Note
The SolidFire driver creates names for volumes on the back end using the format UUID-<cinder-id>. This works well, but there is a possibility of a UUID collision for customers running multiple clouds against the same cluster. In Mitaka the ability was added to eliminate the possibility of collisions by introducing the sf_volume_prefix configuration variable. On the SolidFire cluster each volume will be labeled with the prefix, providing the ability to configure unique volume names for each cloud. The default prefix is ‘UUID-‘.
Changing the setting on an existing deployment will result in the existing volumes being inaccessible. To introduce this change to an existing deployment it is recommended to add the Cluster as if it were a second backend and disable new deployments to the current back end.
| Configuration option = Default value | Description |
|---|---|
| [DEFAULT] | |
sf_account_prefix = None |
(String) Create SolidFire accounts with this prefix. Any string can be used here, but the string “hostname” is special and will create a prefix using the cinder node hostname (previous default behavior). The default is NO prefix. |
sf_allow_template_caching = True |
(Boolean) Create an internal cache of copy of images when a bootable volume is created to eliminate fetch from glance and qemu-conversion on subsequent calls. |
sf_allow_tenant_qos = False |
(Boolean) Allow tenants to specify QOS on create |
sf_api_port = 443 |
(Port number) SolidFire API port. Useful if the device api is behind a proxy on a different port. |
sf_emulate_512 = True |
(Boolean) Set 512 byte emulation on volume creation; |
sf_enable_vag = False |
(Boolean) Utilize volume access groups on a per-tenant basis. |
sf_enable_volume_mapping = True |
(Boolean) Create an internal mapping of volume IDs and account. Optimizes lookups and performance at the expense of memory, very large deployments may want to consider setting to False. |
sf_svip = None |
(String) Overrides default cluster SVIP with the one specified. This is required or deployments that have implemented the use of VLANs for iSCSI networks in their cloud. |
sf_template_account_name = openstack-vtemplate |
(String) Account name on the SolidFire Cluster to use as owner of template/cache volumes (created if does not exist). |
sf_volume_prefix = UUID- |
(String) Create SolidFire volumes with this prefix. Volume names are of the form <sf_volume_prefix><cinder-volume-id>. The default is to use a prefix of ‘UUID-‘. |
Supported operations¶
- Create, delete, attach, and detach volumes.
- Create, list, and delete volume snapshots.
- Create a volume from a snapshot.
- Copy an image to a volume.
- Copy a volume to an image.
- Clone a volume.
- Extend a volume.
- Retype a volume.
- Manage and unmanage a volume.
- Consistency group snapshots.
QoS support for the SolidFire drivers includes the ability to set the
following capabilities in the OpenStack Block Storage API
cinder.api.contrib.qos_specs_manage qos specs extension module:
- minIOPS - The minimum number of IOPS guaranteed for this volume. Default = 100.
- maxIOPS - The maximum number of IOPS allowed for this volume. Default = 15,000.
- burstIOPS - The maximum number of IOPS allowed over a short period of time. Default = 15,000.
- scaledIOPS - The presence of this key is a flag indicating that the above IOPS should be scaled by the following scale values. It is recommended to set the value of scaledIOPS to True, but any value will work. The absence of this key implies false.
- scaleMin - The amount to scale the minIOPS by for every 1GB of additional volume size. The value must be an integer.
- scaleMax - The amount to scale the maxIOPS by for every 1GB of additional volume size. The value must be an integer.
- scaleBurst - The amount to scale the burstIOPS by for every 1GB of additional volume size. The value must be an integer.
The QoS keys above no longer require to be scoped but must be created and associated to a volume type. For information about how to set the key-value pairs and associate them with a volume type, see the volume qos section in the OpenStackClient command list.
Note
When using scaledIOPS, the scale values must be chosen such that the constraint minIOPS <= maxIOPS <= burstIOPS is always true. The driver will enforce this constraint.
Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.