The NetApp unified driver for ONTAP with NFS is a driver interface from OpenStack block storage to a ONTAP cluster system to accomplish provisioning and management of OpenStack volumes on NFS exports provided by the ONTAP cluster system. The NetApp unified driver for the ONTAP cluster does not require any additional management software to achieve the desired functionality. It uses NetApp APIs to interact with the ONTAP cluster.
To set up the NetApp ONTAP NFS driver for Cinder, the
following stanza should be added to the Cinder configuration file
(cinder.conf):
[myNfsBackend]
volume_backend_name=myNfsBackend
volume_driver=cinder.volume.drivers.netapp.common.NetAppDriver
netapp_server_hostname=hostname
netapp_server_port=80
netapp_storage_protocol=nfs
netapp_storage_family=ontap_cluster
netapp_login=admin_username
netapp_password=admin_password
netapp_vserver=svm_name
nfs_shares_config=path_to_nfs_exports_file
max_over_subscription_ratio=1.0
reserved_percentage=5
nfs_mount_options=lookupcache=pos
Be sure that the value of the enabled_backends option in the
[DEFAULT] stanza includes the name of the stanza you chose for
the backend.
Caution
Please note that exported qtrees are not supported by Cinder.
Note
The file referenced in the nfs_shares_config configuration
option should contain the NFS exports in the ip:/share format,
for example:
10.63.165.215:/nfs/test
10.63.165.215:/nfs2/test2
where ip corresponds to the IP address assigned to a Data LIF,
and share refers to a junction path for a FlexVol volume within
an SVM. Make sure that volumes corresponding to exports have
read/write permissions set on the ONTAP controllers. Do not
put mount options in the nfs_shares_config file; use
nfs_mount_options instead. For more information on that and
other parameters available to affect the behavior of NetApp’s NFS
driver, please refer to
http://docs.openstack.org/trunk/config-reference/content/nfs-driver-options.html.
Table 4.20, “Configuration options for ONTAP with NFS” below lists the configuration options available for the unified driver for an ONTAP deployment that uses the NFS storage protocol.
Option |
Type |
Default Value |
Description |
|---|---|---|---|
|
Required |
The hostname or IP address for the storage system or proxy server. The value of this option should be the IP address of either the cluster management LIF or the SVM management LIF. |
|
|
Optional |
The TCP port to use for communication with the storage system or proxy server. If not specified, ONTAP drivers will use 80 for HTTP and 443 for HTTPS. |
|
|
Required |
Administrative user account name used to access the storage system or proxy server. |
|
|
Required |
Password for the administrative user account specified in the |
|
|
Required |
The storage protocol to be used. Valid options are |
|
|
Required |
|
Transport protocol for communicating with the storage system or proxy server. Valid options include |
|
Optional |
This option specifies the path of the NetApp copy offload tool binary. Ensure that the binary has execute permissions set which allow the effective user of the |
|
|
Required |
This option specifies the storage virtual machine (previously called a Vserver) name on the storage cluster on which provisioning of block storage volumes should occur. |
|
|
Optional |
|
The storage family type used on the storage system; valid values are |
|
Required |
|
The file referenced by this configuration option should contain a list of NFS shares, each on their own line, to which the driver should attempt to provision new Cinder volumes into. |
|
Optional |
This option is a comma-separated list of options (valid values include |
|
|
Optional |
|
A regular expression to limit the API tracing. This option is honored only if enabling |
|
Optional |
|
Controls whether Cinder volumes backed by NFS backends are sparsely or thickly provisioned. True == sparse, False == Thick. |
|
Optional |
None |
Mount options passed to the nfs client. See section of the nfs man page for details. |
|
Optional |
|
If ‘false’, backing files for cinder volumes are readable by owner, group, and world; if ‘true’, only by owner and group. If ‘auto’ and there are existing Cinder volumes, value will be set to ‘false’ (for backwards compatibility); if ‘auto’ and there are no existing Cinder volumes, the value will be set to ‘true’. |
|
Optional |
|
If ‘false’, operations on the backing files run as root; if ‘true’, operations on the backing files for cinder volumes run unprivileged, as the cinder user, and are allowed to succeed even when root is squashed. If ‘auto’ and there are existing Cinder volumes, value will be set to ‘false’ (for backwards compatibility); if ‘auto’ and there are no existing Cinder volumes, the value will be set to ‘true’. |
|
Optional |
|
If the percentage of available space for an NFS share has dropped below the value specified by this option, the NFS image cache will be cleaned. |
|
Optional |
|
When the percentage of available space on an NFS share has reached the percentage specified by this option, the driver will stop clearing files from the NFS image cache that have not been accessed in the last M minutes, where M is the value of the |
|
Optional |
|
This option specifies the threshold for last access time for images in the NFS image cache. When a cache cleaning cycle begins, images in the cache that have not been accessed in the last M minutes, where M is the value of this parameter, will be deleted from the cache to create free space on the NFS share. |
|
Optional |
|
This option represents the amount of total capacity of a storage pool that will be reserved and cannot be utilized for provisioning Cinder volumes. |
|
Optional |
|
This option is defined as a float, and specifies the amount of over-provisioning to allow when thin provisioning is being used in the storage pool. A value of 1.0 will mean that the provisioned capacity will not be able to exceed the total capacity, while larger values will result in increased levels of allowed over-provisioning. |
|
Optional |
(see description) |
This option may be used to override the default filter function, which prevents Cinder from placing new volumes on storage controllers that may become overutilized. The default value is “capabilities.utilization < 70”. |
|
Optional |
(see description) |
This option may be used to override the default goodness function, which allows Cinder to place new volumes on lesser-utilized storage controllers. The default value is “100 - capabilities.utilization”. |
Table 4.20. Configuration options for ONTAP with NFS
Caution
If you specify an account in the netapp_login option that only
has SVM administration privileges (rather than cluster
administration privileges), some advanced features of the NetApp
unified driver will not work and you may see warnings in the Cinder
logs. See the section called “Account Permission Considerations”
for more details on the required access level permissions for an SVM
admin account.
Important
The NFS client cache refresh interval can vary depending on how the
NFS client’s default mounting options are configured. In order to
prevent the issue of a stale negative cache
entry, an additional option must be passed to the NFS mount command
invoked by the Cinder using an NFS driver. This can be configured by
adding the line nfs_mount_options = lookupcache=pos to your
driver configuration stanza in your cinder.conf file. Alternatively,
if you are already setting other NFS mount options, then you can
just add ,lookupcache=pos to the end of your current
“nfs_mount_options”. The effect of this additional option is to
force the NFS client to ignore any negative entries in its cache and
always check the NFS host when attempting to confirm the existence
of a file.
Please be aware, the nfs_mount_options values are not
automatically applied if the export is already mounted. In such
a case, the nfs_mount_options values are applied the next
time the export is mounted.
This document is licensed under Apache 2.0 license.