This OpenStack Block Storage volume driver provides iSCSI and NFS support for Hitachi NAS Platform Models 3080, 3090, 4040, 4060, 4080, and 4100.
The NFS and iSCSI drivers support these operations:
Before using iSCSI and NFS services, use the HNAS configuration and management GUI (SMU) or SSC CLI to create storage pool(s), file system(s), and assign an EVS. Make sure that the file system used is not created as a replication target. Additionally:
Create NFS exports, choose a path for them (it must be different from /) and set the Show snapshots option to hide and disable access.
Also, in the Access Configuration set the option norootsquash, For example, "* (rw, norootsquash)", so HNAS cinder driver can change the permissions of its volumes.
In order to use the hardware accelerated features of NFS HNAS, we recommend setting max-nfs-version to 3. Refer to the HNAS command-line reference to see how to configure this option.
The HNAS driver is supported for Red Hat Enterprise Linux OpenStack Platform, SUSE OpenStack Cloud, and Ubuntu OpenStack. The following packages must be installed:
If you are not using SSH, you need the HDS SSC to communicate with an HNAS array using the SSC commands. This utility package is available in the RPM package distributed with the hardware through physical media or it can be manually copied from the SMU to the Block Storage host.
If you are installing the driver from an RPM or DEB package, follow the steps below:
Install the dependencies:
In Red Hat:
# yum install nfs-utils nfs-utils-lib
Or in Ubuntu:
# apt-get install nfs-common
Or in SUSE:
# zypper install nfs-client
If you are using Ubuntu 12.04, you also need to install libc6-i386
# apt-get install libc6-i386
Configure the driver as described in the Driver configuration section.
Restart all cinder services (volume, scheduler and backup).
The HDS driver supports the concept of differentiated services (also referred as quality of service) by mapping volume types to services provided through HNAS.
HNAS supports a variety of storage options and file system capabilities, which are selected through the definition of volume types and the use of multiple back ends. The driver maps up to four volume types into separated exports or file systems, and can support any number if using multiple back ends.
The configuration for the driver is read from an XML-formatted file (one per back end), which you need to create and set its path in the cinder.conf configuration file. Below are the settings needed in the cinder.conf configuration file [1]:
[DEFAULT]
enabled_backends = hnas_iscsi1, hnas_nfs1
For HNAS iSCSI driver create this section:
[hnas_iscsi1]
volume_driver = cinder.volume.drivers.hitachi.hnas_iscsi.HDSISCSIDriver
hds_hnas_iscsi_config_file = /path/to/config/hnas_config_file.xml
volume_backend_name = HNAS-ISCSI
For HNAS NFS driver create this section:
[hnas_nfs1]
volume_driver = cinder.volume.drivers.hitachi.hnas_nfs.HDSNFSDriver
hds_hnas_nfs_config_file = /path/to/config/hnas_config_file.xml
volume_backend_name = HNAS-NFS
The XML file has the following format:
<?xml version = "1.0" encoding = "UTF-8" ?>
<config>
<mgmt_ip0>172.24.44.15</mgmt_ip0>
<hnas_cmd>ssc</hnas_cmd>
<chap_enabled>False</chap_enabled>
<ssh_enabled>False</ssh_enabled>
<cluster_admin_ip0>10.1.1.1</cluster_admin_ip0>
<username>supervisor</username>
<password>supervisor</password>
<svc_0>
<volume_type>default</volume_type>
<iscsi_ip>172.24.44.20</iscsi_ip>
<hdp>fs01-husvm</hdp>
</svc_0>
<svc_1>
<volume_type>platinum</volume_type>
<iscsi_ip>172.24.44.20</iscsi_ip>
<hdp>fs01-platinum</hdp>
</svc_1>
</config>
An OpenStack Block Storage node using HNAS drivers can have up to four services. Each service is defined by a svc_n tag (svc_0, svc_1, svc_2, or svc_3 [2], for example). These are the configuration options available for each service label:
Option | Type | Default | Description |
---|---|---|---|
volume_type | Required | default | When a create_volume call with a certain volume type happens, the volume type will try to be matched up with this tag. In each configuration file you must define the default volume type in the service labels and, if no volume type is specified, the default is used. Other labels are case sensitive and should match exactly. If no configured volume types match the incoming requested type, an error occurs in the volume creation. |
iscsi_ip | Required only for iSCSI | An iSCSI IP address dedicated to the service. | |
hdp | Required | For iSCSI driver: virtual file system label associated with the service. For NFS driver: path to the volume (<ip_address>:/<path>) associated with the service. Additionally, this entry must be added in the file used to list available NFS shares. This file is located, by default, in /etc/cinder/nfs_shares or you can specify the location in the nfs_shares_config option in the cinder.conf configuration file. |
These are the configuration options available to the config section of the XML configuration file:
Option | Type | Default | Description |
---|---|---|---|
mgmt_ip0 | Required | Management Port 0 IP address. Should be the IP address of the Admin EVS. | |
hnas_cmd | Optional | ssc | Command to communicate to HNAS array. |
chap_enabled | Optional (iSCSI only) | True | Boolean tag used to enable CHAP authentication protocol. |
username | Required | supervisor | User name is always required on HNAS. |
password | Required | supervisor | Password is always required on HNAS. |
svc_0, svc_1, svc_2, svc_3 | Optional | (at least one label has to be defined) | Service labels: these four predefined names help four different sets of configuration options. Each can specify HDP and a unique volume type. |
cluster_admin_ip0 | Optional if ssh_enabled is True | The address of HNAS cluster admin. | |
ssh_enabled | Optional | False | Enables SSH authentication between Block Storage host and the SMU. |
ssh_private_key | Required if ssh_enabled is True | False | Path to the SSH private key used to authenticate in HNAS SMU. The public key must be uploaded to HNAS SMU using ssh-register-public-key (this is an SSH subcommand). Note that copying the public key HNAS using ssh-copy-id does not work properly as the SMU periodically wipe out those keys. |
HNAS driver supports differentiated types of service using the service labels. It is possible to create up to four types of them, as gold, platinum, silver, and ssd, for example.
After creating the services in the XML configuration file, you must configure one volume_type per service. Each volume_type must have the metadata service_label with the same name configured in the <volume_type> section of that service. If this is not set, the Block Storage service will schedule the volume creation to the pool with largest available free space or other criteria configured in volume filters.
$ cinder type-create default
$ cinder type-key default set service_label=default
$ cinder type-create platinum-tier
$ cinder type-key platinum set service_label=platinum
If you use multiple back ends and intend to enable the creation of a volume in a specific back end, you must configure volume types to set the volume_backend_name option to the appropriate back end. Then, create volume_type configurations with the same volume_backend_name.
$ cinder type-create 'iscsi'
$ cinder type-key 'iscsi' set volume_backend_name = 'HNAS-ISCSI'
$ cinder type-create 'nfs'
$ cinder type-key 'nfs' set volume_backend_name = 'HNAS-NFS'
You can deploy multiple OpenStack HNAS driver instances that each control a separate HNAS array. Each service (svc_0, svc_1, svc_2, svc_3) on the instance need to have a volume_type and service_label metadata associated with it. If no metadata is associated with a pool, the Block Storage filtering algorithm selects the pool with the largest available free space.
Instead of using SSC commands on the Block Storage host and storing its credentials in the XML configuration file, the HNAS driver supports SSH authentication. To configure that:
If you don’t have a pair of public keys already generated, create one on the Block Storage host (leave the pass-phrase empty):
$ mkdir -p /opt/hds/ssh
$ ssh-keygen -f /opt/hds/ssh/hnaskey
Change the owner of the key to cinder (or the user under which the volume service will be run):
# chown -R cinder.cinder /opt/hds/ssh
Create the directory ssh_keys in the SMU server:
$ ssh [manager|supervisor]@<smu-ip> 'mkdir -p /var/opt/mercury-main/home/[manager|supervisor]/ssh_keys/'
Copy the public key to the ssh_keys directory:
$ scp /opt/hds/ssh/hnaskey.pub [manager|supervisor]@<smu-ip>:/var/opt/mercury-main/home/[manager|supervisor]/ssh_keys/
Access the SMU server:
$ ssh [manager|supervisor]@<smu-ip>
Run the command to register the SSH keys:
$ ssh-register-public-key -u [manager|supervisor] -f ssh_keys/hnaskey.pub
Check the communication with HNAS on the Block Storage host:
$ ssh -i /opt/hds/ssh/hnaskey [manager|supervisor]@<smu-ip> 'ssc <cluster_admin_ip0> df -a'
<cluster_admin_ip0> is localhost for single node deployments. This should return a list of available file systems on HNAS.
Warning
Note that copying the public key HNAS using ssh-copy-id does not work properly as the SMU periodically wipes out those keys.
Manage and unmanage are two new API extensions that add some new features to the driver. The manage action on an existing volume is very similar to a volume creation. It creates a volume entry in the Block Storage database, but instead of creating a new volume in the back end, it only adds a link to an existing volume. Volume name, description, volume_type, metadata, and availability_zone are supported as in a normal volume creation.
The unmanage action on an existing volume removes the volume from the Block Storage database, but keeps the actual volume in the back-end. From a Block Storage perspective the volume would be deleted, but it would still exist for outside use.
On the Dashboard:
For NFS:
For iSCSI:
By CLI:
$ cinder --os-volume-api-version 2 manage [--source-name <source-name>][--id-type <id-type>]
[--name <name>][--description <description>][--volume-type <volume-type>]
[--availability-zone <availability-zone>][--metadata [<key=value> [<key=value> ...]]][--bootable]
<host> [<key=value> [<key=value> ...]]
Example:
For NFS:
$ cinder --os-volume-api-version 2 manage --name <volume-test> --volume-type <silver>
--source-name <172.24.44.34:/silver/volume-test> <ubuntu@hnas-nfs#test_silver>
For iSCSI:
$ cinder --os-volume-api-version 2 manage --name <volume-test> --volume-type <silver>
--source-name <filesystem-test/volume-test> <ubuntu@hnas-iscsi#test_silver>
On the Dashboard:
By CLI:
$ cinder --os-volume-api-version 2 unmanage <volume>
Example:
$ cinder --os-volume-api-version 2 unmanage <voltest>
The get_volume_stats() function always provides the available capacity based on the combined sum of all the HDPs that are used in these services labels.
After changing the configuration on the storage node, the Block Storage driver must be restarted.
On Red Hat, if the system is configured to use SELinux, you need to set virt_use_nfs = on for NFS driver work properly.
# setsebool -P virt_use_nfs on
It is not possible to manage a volume if there is a slash (‘/’) or a colon (‘:’) in the volume name.
Footnotes
[1] | The configuration file location may differ. |
[2] | There is no relative precedence or weight among these four labels. |
Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.