The volume management driver for Storwize family and SAN Volume Controller (SVC) provides OpenStack Compute instances with access to IBM Storwize family or SVC storage systems.
Storwize/SVC driver supports the following Block Storage service volume operations:
The Storwize family or SVC system must be configured for iSCSI, Fibre Channel, or both.
If using iSCSI, each Storwize family or SVC node should have at least one iSCSI IP address. The IBM Storwize/SVC driver uses an iSCSI IP address associated with the volume’s preferred node (if available) to attach the volume to the instance, otherwise it uses the first available iSCSI IP address of the system. The driver obtains the iSCSI IP address directly from the storage system. You do not need to provide these iSCSI IP addresses directly to the driver.
Note
If using iSCSI, ensure that the compute nodes have iSCSI network access to the Storwize family or SVC system.
If using Fibre Channel (FC), each Storwize family or SVC node should have at least one WWPN port configured. The driver uses all available WWPNs to attach the volume to the instance. The driver obtains the WWPNs directly from the storage system. You do not need to provide these WWPNs directly to the driver.
Note
If using FC, ensure that the compute nodes have FC connectivity to the Storwize family or SVC system.
If using iSCSI for data access and the
storwize_svc_iscsi_chap_enabled
is set to True
, the driver will
associate randomly-generated CHAP secrets with all hosts on the Storwize
family system. The compute nodes use these secrets when creating
iSCSI connections.
Warning
CHAP secrets are added to existing hosts as well as newly-created ones. If the CHAP option is enabled, hosts will not be able to access the storage without the generated secrets.
Note
Not all OpenStack Compute drivers support CHAP authentication. Please check compatibility before using.
Note
CHAP secrets are passed from OpenStack Block Storage to Compute in clear text. This communication should be secured to ensure that CHAP secrets are not discovered.
The IBM Storwize/SVC driver can allocate volumes in multiple pools.
The pools should be created in advance and be provided to the driver
using the storwize_svc_volpool_name
configuration flag in the form
of a comma-separated list.
For the complete list of configuration flags, see Storwize family and SVC driver options in cinder.conf.
The driver requires access to the Storwize family or SVC system
management interface. The driver communicates with the management using
SSH. The driver should be provided with the Storwize family or SVC
management IP using the san_ip
flag, and the management port should
be provided by the san_ssh_port
flag. By default, the port value is
configured to be port 22 (SSH). Also, you can set the secondary
management IP using the storwize_san_secondary_ip
flag.
Note
Make sure the compute node running the cinder-volume management driver has SSH network access to the storage system.
To allow the driver to communicate with the Storwize family or SVC system, you must provide the driver with a user on the storage system. The driver has two authentication methods: password-based authentication and SSH key pair authentication. The user should have an Administrator role. It is suggested to create a new user for the management driver. Please consult with your storage and security administrator regarding the preferred authentication method and how passwords or SSH keys should be stored in a secure manner.
Note
When creating a new user on the Storwize or SVC system, make sure the user belongs to the Administrator group or to another group that has an Administrator role.
If using password authentication, assign a password to the user on the
Storwize or SVC system. The driver configuration flags for the user and
password are san_login
and san_password
, respectively.
If you are using the SSH key pair authentication, create SSH private and
public keys using the instructions below or by any other method.
Associate the public key with the user by uploading the public key:
select the choose file option in the Storwize family or SVC
management GUI under SSH public key. Alternatively, you may
associate the SSH public key using the command-line interface; details can
be found in the Storwize and SVC documentation. The private key should be
provided to the driver using the san_private_key
configuration flag.
You can create an SSH key pair using OpenSSH, by running:
$ ssh-keygen -t rsa
The command prompts for a file to save the key pair. For example, if you
select key
as the filename, two files are created: key
and
key.pub
. The key
file holds the private SSH key and key.pub
holds the public SSH key.
The command also prompts for a pass phrase, which should be empty.
The private key file should be provided to the driver using the
san_private_key
configuration flag. The public key should be
uploaded to the Storwize family or SVC system using the storage
management GUI or command-line interface.
Note
Ensure that Cinder has read permissions on the private key file.
Set the volume driver to the Storwize family and SVC driver by setting
the volume_driver
option in the cinder.conf
file as follows:
iSCSI:
[svc1234]
volume_driver = cinder.volume.drivers.ibm.storwize_svc.storwize_svc_iscsi.StorwizeSVCISCSIDriver
san_ip = 1.2.3.4
san_login = superuser
san_password = passw0rd
storwize_svc_volpool_name = cinder_pool1
volume_backend_name = svc1234
FC:
[svc1234]
volume_driver = cinder.volume.drivers.ibm.storwize_svc.storwize_svc_fc.StorwizeSVCFCDriver
san_ip = 1.2.3.4
san_login = superuser
san_password = passw0rd
storwize_svc_volpool_name = cinder_pool1
volume_backend_name = svc1234
Add the following to the back-end specification to specify another storage to replicate to:
replication_device = backend_id:rep_svc,
san_ip:1.2.3.5,
san_login:superuser,
san_password:passw0rd,
pool_name:cinder_pool1
The backend_id
is a unique name of the remote storage, the san_ip
,
san_login
, and san_password
is authentication information for the
remote storage. The pool_name
is the pool name for the replication
target volume.
Note
Only one replication_device
can be configured for one back end
storage since only one replication target is supported now.
The following options specify default values for all volumes. Some can be over-ridden using volume types, which are described below.
Configuration option = Default value | Description |
---|---|
[DEFAULT] | |
san_ip = |
(String) IP address of SAN controller |
san_login = admin |
(String) Username for SAN controller |
san_password = |
(String) Password for SAN controller |
san_private_key = |
(String) Filename of private key to use for SSH authentication |
san_ssh_port = 22 |
(Port number) SSH port to use with SAN |
storwize_san_secondary_ip = None |
(String) Specifies secondary management IP or hostname to be used if san_ip is invalid or becomes inaccessible. |
storwize_svc_allow_tenant_qos = False |
(Boolean) Allow tenants to specify QOS on create |
storwize_svc_flashcopy_rate = 50 |
(Integer) Specifies the Storwize FlashCopy copy rate to be used when creating a full volume copy. The default is rate is 50, and the valid rates are 1-100. |
storwize_svc_flashcopy_timeout = 120 |
(Integer) Maximum number of seconds to wait for FlashCopy to be prepared. |
storwize_svc_iscsi_chap_enabled = True |
(Boolean) Configure CHAP authentication for iSCSI connections (Default: Enabled) |
storwize_svc_multihostmap_enabled = True |
(Boolean) DEPRECATED: This option no longer has any affect. It is deprecated and will be removed in the next release. |
storwize_svc_multipath_enabled = False |
(Boolean) Connect with multipath (FC only; iSCSI multipath is controlled by Nova) |
storwize_svc_stretched_cluster_partner = None |
(String) If operating in stretched cluster mode, specify the name of the pool in which mirrored copies are stored.Example: “pool2” |
storwize_svc_vol_autoexpand = True |
(Boolean) Storage system autoexpand parameter for volumes (True/False) |
storwize_svc_vol_compression = False |
(Boolean) Storage system compression option for volumes |
storwize_svc_vol_easytier = True |
(Boolean) Enable Easy Tier for volumes |
storwize_svc_vol_grainsize = 256 |
(Integer) Storage system grain size parameter for volumes (32/64/128/256) |
storwize_svc_vol_iogrp = 0 |
(Integer) The I/O group in which to allocate volumes |
storwize_svc_vol_nofmtdisk = False |
(Boolean) Specifies that the volume not be formatted during creation. |
storwize_svc_vol_rsize = 2 |
(Integer) Storage system space-efficiency parameter for volumes (percentage) |
storwize_svc_vol_warning = 0 |
(Integer) Storage system threshold for volume capacity warnings (percentage) |
storwize_svc_volpool_name = volpool |
(List) Comma separated list of storage system storage pools for volumes. |
Note the following:
san_password
) or
SSH private key (san_private_key
). One must be specified. If
both are specified, the driver uses only the SSH private key.storwize_svc_vol_rsize
flag defines the initial physical
allocation percentage for thin-provisioned volumes, or if set to
-1
, the driver creates full allocated volumes. More details about
the available options are available in the Storwize family and SVC
documentation.The IBM Storwize/SVC driver exposes capabilities that can be added to
the extra specs
of volume types, and used by the filter
scheduler to determine placement of new volumes. Make sure to prefix
these keys with capabilities:
to indicate that the scheduler should
use them. The following extra specs
are supported:
capabilities:volume_backend_name
- Specify a specific back-end
where the volume should be created. The back-end name is a
concatenation of the name of the IBM Storwize/SVC storage system as
shown in lssystem
, an underscore, and the name of the pool (mdisk
group). For example:
capabilities:volume_backend_name=myV7000_openstackpool
capabilities:compression_support
- Specify a back-end according to
compression support. A value of True
should be used to request a
back-end that supports compression, and a value of False
will
request a back-end that does not support compression. If you do not
have constraints on compression support, do not set this key. Note
that specifying True
does not enable compression; it only
requests that the volume be placed on a back-end that supports
compression. Example syntax:
capabilities:compression_support='<is> True'
capabilities:easytier_support
- Similar semantics as the
compression_support
key, but for specifying according to support
of the Easy Tier feature. Example syntax:
capabilities:easytier_support='<is> True'
capabilities:pool_name
- Specify a specific pool to create volume
if only multiple pools are configured. pool_name should be one value
configured in storwize_svc_volpool_name flag. Example syntax:
capabilities:pool_name=cinder_pool2
Volume types can also be used to pass options to the IBM Storwize/SVC
driver, which over-ride the default values set in the configuration
file. Contrary to the previous examples where the capabilities
scope
was used to pass parameters to the Cinder scheduler, options can be
passed to the IBM Storwize/SVC driver with the drivers
scope.
The following extra specs
keys are supported by the IBM Storwize/SVC
driver:
These keys have the same semantics as their counterparts in the
configuration file. They are set similarly; for example, rsize=2
or
compression=False
.
In the following example, we create a volume type to specify a controller that supports compression, and enable compression:
$ openstack volume type create compressed
$ openstack volume type set --property capabilities:compression_support='<is> True' --property drivers:compression=True compressed
We can then create a 50GB volume using this type:
$ openstack volume create "compressed volume" --type compressed --size 50
In the following example, create a volume type that enables synchronous replication (metro mirror):
$ openstack volume type create ReplicationType
$ openstack volume type set --property replication_type="<in> metro" \
--property replication_enabled='<is> True' --property volume_backend_name=svc234 ReplicationType
In the following example, we create a volume type to support stretch cluster volume or mirror volume:
$ openstack volume type create mirror_vol_type
$ openstack volume type set --property volume_backend_name=svc1 \
--property drivers:mirror_pool=pool2 mirror_vol_type
Volume types can be used, for example, to provide users with different
The Storwize driver provides QOS support for storage volumes by
controlling the I/O amount. QOS is enabled by editing the
etc/cinder/cinder.conf
file and setting the
storwize_svc_allow_tenant_qos
to True
.
There are three ways to set the Storwize IOThrotting
parameter for
storage volumes:
qos:IOThrottling
key into a QOS specification and
associate it with a volume type.qos:IOThrottling
key into an extra specification with a
volume type.qos:IOThrottling
key to the storage volume metadata.Note
If you are changing a volume type with QOS to a new volume type without QOS, the QOS configuration settings will be removed.
In the context of OpenStack Block Storage’s volume migration feature, the IBM Storwize/SVC driver enables the storage’s virtualization technology. When migrating a volume from one pool to another, the volume will appear in the destination pool almost immediately, while the storage moves the data in the background.
Note
To enable this feature, both pools involved in a given volume
migration must have the same values for extent_size
. If the
pools have different values for extent_size
, the data will still
be moved directly between the pools (not host-side copy), but the
operation will be synchronous.
The IBM Storwize/SVC driver allows for extending a volume’s size, but only for volumes without snapshots.
Snapshots are implemented using FlashCopy with no background copy (space-efficient). Volume clones (volumes created from existing volumes) are implemented with FlashCopy, but with background copy enabled. This means that volume clones are independent, full copies. While this background copy is taking place, attempting to delete or extend the source volume will result in that operation waiting for the copy to complete.
The IBM Storwize/SVC driver enables you to modify volume types. When you modify volume types, you can also change these extra specs properties:
Note
When you change the rsize
, grainsize
or compression
properties, volume copies are asynchronously synchronized on the
array.
Note
To change the iogrp
property, IBM Storwize/SVC firmware version
6.4.0 or later is required.
A volume is only replicated if the volume is created with a volume-type
that has the extra spec replication_enabled
set to <is> True
. Three
types of replication are supported now, global mirror(async), global mirror
with change volume(async) and metro mirror(sync). It can be specified by a
volume-type that has the extra spec replication_type
set to
<in> global
, <in> gmcv
or <in> metro
. If no replication_type
is specified, global mirror will be created for replication.
If replication_type
set to <in> gmcv
, cycle_period_seconds can be
set as the cycling time perform global mirror relationship with multi cycling
mode. Default value is 300. Example syntax:
$ cinder type-create gmcv_type
$ cinder type-key gmcv_type set replication_enabled='<is> True' \
replication_type="<in> gmcv" drivers:cycle_period_seconds=500
Note
It is better to establish the partnership relationship between the replication source storage and the replication target storage manually on the storage back end before replication volume creation.
The failover-host
command is designed for the case where the primary
storage is down.
$ cinder failover-host cinder@svciscsi --backend_id target_svc_id
If a failover command has been executed and the primary storage has
been restored, it is possible to do a failback by simply specifying
default as the backend_id
:
$ cinder failover-host cinder@svciscsi --backend_id default
Note
Before you perform a failback operation, synchronize the data from the replication target volume to the primary one on the storage back end manually, and do the failback only after the synchronization is done since the synchronization may take a long time. If the synchronization is not done manually, Storwize Block Storage service driver will perform the synchronization and do the failback after the synchronization is finished.
Before creating replication group, a group-spec which key
consistent_group_replication_enabled
set to <is> True
should be
set in group type. Volume type used to create group must be replication
enabled, and its replication_type
should be set either <in> global
or <in> metro
. The “failover_group” api allows group to be failed over
and back without failing over the entire host. Example syntax:
$ cinder group-type-create rep-group-type-example
$ cinder group-type-key rep-group-type-example set consistent_group_replication_enabled='<is> True'
$ cinder type-create type-global
$ cinder type-key type-global set replication_enabled='<is> True' replication_type='<in> global'
$ cinder group-create rep-group-type-example type-global --name global-group
$ cinder group-failover-replication --secondary-backend-id target_svc_id group_id
$ cinder group-failover-replication --secondary-backend-id default group_id
Note
Option allow-attached-volume can be used to failover the in-use volume, but fail over/back an in-use volume is not recommended. If the user does failover operation to an in-use volume, the volume status remains in-use after failover. But the in-use replication volume would change to read-only since the primary volume is changed to auxiliary side and the instance is still attached to the master volume. As a result please detach the replication volume first and attach again if user want to reuse the in-use replication volume as read-write.
A hyperswap volume is created with a volume-type that has the extra spec
drivers:volume_topology
set to hyperswap
.
To support hyperswap volumes, IBM Storwize/SVC firmware version 7.6.0 or
later is required.
Add the following to the back-end configuration to specify the host preferred
site for hyperswap volume.
FC:
storwize_preferred_host_site = site1:20000090fa17311e&ff00000000000001,
site2:20000089762sedce&ff00000000000000
iSCSI:
storwize_preferred_host_site = site1:iqn.1993-08.org.debian:01:eac5ccc1aaa&iqn.1993-08.org.debian:01:be53b7e236be,
site2:iqn.1993-08.org.debian:01:eac5ccc1bbb&iqn.1993-08.org.debian:01:abcdefg9876w
The site1 and site2 are names of the two host sites used in Storwize storage. The WWPNs and IQNs are the connectors used for host mapping in Storwize.
$ cinder type-create hyper_type
$ cinder type-key hyper_type set drivers:volume_topology=hyperswap \
drivers:peer_pool=Pool_site2
Note
The property rsize
is considered as buffersize
for hyperswap
volume.
The hyperswap property iogrp
is selected by storage.
A group is created as a hyperswap group with a group-type that has the
group spec hyperswap_group_enabled
set to <is> True
.
Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.