29 KiB

Raw Blame History

IBM Storage Driver for OpenStack

Introduction

The IBM Storage Driver for OpenStack is a software component of the OpenStack cloud environment that enables utilization of storage resources provided by supported IBM storage systems.

The driver was validated on storage systems, as detailed in the Supported storage systems section below.

After the driver is configured on the OpenStack Cinder nodes, storage volumes can be allocated by the Cinder nodes to the Nova nodes. Virtual machines on the Nova nodes can then utilize these storage resources.

Note

Unless stated otherwise, all references to XIV storage system in this guide relate all members of the Spectrum Accelerate Family (SAF): XIV, Spectrum Accelerate, FlashSystem A9000/A9000R.

Concept diagram

This figure illustrates how an IBM storage system is connected to the OpenStack cloud environment and provides storage resources when the IBM Storage Driver for OpenStack is configured on the OpenStack Cinder nodes. The OpenStack cloud is connected to the IBM storage system over Fibre Channel or iSCSI (DS8000 systems support only Fibre Channel connections). Remote cloud users can issue requests for storage resources from the OpenStack cloud. These requests are transparently handled by the IBM Storage Driver, which communicates with the IBM storage system and controls the storage volumes on it. The IBM storage resources are then provided to the Nova nodes in the OpenStack cloud.

Preparation

If you intend to manage a Spectrum Accelerate Family product, you need to install a Python client for executing CLI commands on all Cinder nodes. The IBM Python XCLI Client allows full management and monitoring of the relevant storage systems.

The client package and its documentation are available at GitHub.

Compatibility and requirements

This section specifies the compatibility and requirements of the IBM Storage Driver for OpenStack.

Supported storage systems

The IBM Storage Driver for OpenStack supports the IBM storage systems, as detailed in the following table.

Storage system	Microcode version	Connectivity
IBM DS8870	7.5 SP4 or later, 7.5 with RESTful API patch	Fibre Channel (FC)
IBM DS8880	8.1 or later	Fibre Channel (FC)
IBM XIV Storage System	11.1.x, 11.2.x, 11.3.x, 11.4.x, 11.5.x, 11.6.x	Fibre Channel (FC) iSCSI
IBM Spectrum Accelerate	11.5.x	iSCSI
IBM FlashSystem A9000/A9000R	12.0.x, 12.1.x, 12.2.x	Fibre Channel (FC) iSCSI

Copy Services license

Copy Services features help you implement storage solutions to keep your business running 24 hours a day, 7 days a week by providing image caching, replication and cloning functions. The Copy Services license is based on usable capacity of the volumes involved in Copy Services functionality.

The Copy Services license is available for the following license scopes: FB and ALL (both FB and CKD).

The Copy Services license includes the following features:

Global Mirror
Metro Mirror
Metro/Global Mirror
Point-in-Time Copy/FlashCopy®
z/OS® Global Mirror
z/OS Metro/Global Mirror Incremental Resync (RMZ)

The Copy Services license feature codes are ordered in increments up to a specific capacity. For example, if you require 160 TB of capacity, order 10 of feature code 8251 (10 TB each up to 100 TB capacity), and 4 of feature code 8252 (15 TB each, for an extra 60 TB).

The Copy Services license includes the following feature codes.

Feature Code	Feature code for licensed function indicator
8250	CS - inactive
8251	CS - 10 TB (up to 100 TB capacity)
8252	CS - 15 TB (from 100.1 TB to 250 TB capacity)
8253	CS - 25 TB (from 250.1 TB to 500 TB capacity)
8254	CS - 75 TB (from 500.1 to 1250 TB capacity)
8255	CS - 175 TB (from 1250.1 TB to 3000 TB capacity)
8256	CS - 300 TB (from 3000.1 TB to 6000 TB capacity)
8260	CS - 500 TB (from 6000.1 TB to 10,000 TB capacity)

The following ordering rules apply when you order the Copy Services license:

The Copy Services license should be ordered based on the total usable capacity of all volumes involved in one or more Copy Services relationships.
The licensed authorization must be equal to or less that the total usable capacity allocated to the volumes that participate in Copy Services operations.
You must purchase features for both the source (primary) and target (secondary) storage system.

Required software on the OpenStack Cinder and Nova nodes

The IBM Storage Driver makes use of the following software on the OpenStack Cinder and Nova-compute nodes.

Software	Installed on
Ubuntu Server (16.04), x64 Red Hat Enterprise Linux (RHEL) 7.x, x64 CentOS Linux 7.x, x64 KVM for IBM z Systems	All OpenStack Cinder nodes
IBM Storage Host Attachment Kit for Linux	All OpenStack Cinder and Nova compute nodes that connect to storage systems and use RHEL 7.x or CentOS Linux 7.x
Linux patch package	All OpenStack Cinder nodes
sysfsutils utility	All OpenStack Cinder nodes on FC network
iscsi-initiator-utils utility (RHEL and CentOS) or open-iscsi utility (Ubuntu)	All OpenStack Cinder and Nova compute nodes on iSCSI network
IBM Python XCLI client	All OpenStack Cinder nodes

Configuration

Configure the driver manually by changing the cinder.conf file as follows:

volume_driver = cinder.volume.drivers.ibm.ibm_storage.IBMStorageDriver

Configuration Description for DS8000

Replication parameters

Parameter	Description	Applicable to
replication _device	Volume replication parameters	DS8000
backend_id	IP address or host name of the target storage system	DS8000
san_login	User name to be used during replication procedure	DS8000
san_password	Password to be used during replication procedure (base64-encoded)	DS8000
san_clustername	Pool name on the target storage system	DS8000
port_pairs	ID pairs of IO ports, participating in replication	DS8000
lss_range_for _cg	LSS range to reserve for consistency groups	DS8000

Configuration Description for SAF

Parameter	Description	Applicable to
management_ips	IP addresses of the management interfaces of a storage system	SAF
san_password	Storage system password (base64-encoded)	SAF
san_login	Storage system user name	SAF
volume_driver	Driver to use for volume creation	SAF
proxy	Proxy for IBM storage driver location within Cinder SAF: cinder.volume.drivers. ibm.ibm_storage.xiv_proxy. XIVProxy DS8000: cinder.volume. drivers.ibm.ibm_storage. xiv_proxy.XIVProxy	SAF
san_ip	Storage system IP address or hostname	SAF
connection_type	Network connection type Values: fibre_channel, iscsi	SAF
san_clustername	Storage pool name	SAF
chap	iSCSI CHAP authentication usage Values: disabled, enabled	SAF
system_id	Storage system ID	SAF

Security

The following information provides an overview of security for the IBM Storage Driver for OpenStack.

Avoiding man-in-the-middle attacks

When using a Spectrum Accelerate Family product, you can prevent man-in-the-middle (MITM) attacks by following these rules:

Upgrade to IBM XIV storage system version 11.3 or later.
If working in a secure mode, do not work insecurely against another storage system in the same environment.
Validate the storage certificate. If you are using an XIV-provided certificate, use the CA file that was provided with your storage system (XIV-CA.pem). The certificate files should be copied to one of the following directories:
- /etc/ssl/certs
- /etc/ssl/certs/xiv
- /etc/pki
- /etc/pki/xiv

If you are using your own certificates, copy them to the same directories with the prefix XIV and in the .pem format. For example: XIV-my_cert.pem.

In order to prevent the CVE-2014-3566 MITM attack, follow these directions.

Configuring Cinder nodes for trusted communication (DS8000 Family)

The IBM Storage Driver for OpenStack communicates with DS8000 over HTTPS, using self-signed certificate or certificate signed by a certificate authority (CA). Configure a trusted communication link to ensure a successful attachment of a Cinder node to a DS8000 storage system, as detailed in the following sections.

Configuring trusted communication link

Before configuring a DS8000 backend, complete the following steps to establish the chain of trust.

In your operating system shell, run this command to obtain the certificate: openssl x509 -in <(openssl s_client -connect <host fqdn>:8452 -prexit 2>/dev/null) -text -out <host fqdn>.pem

If the certificate is self-signed, the following information is displayed:
```
---
Certificate chain
0 s:/CN=ds8000.ibm.com
i:/CN=ds8000.ibm.com
---
```
Create an exception by moving the certificate <fqdn>.pem to the /opt/ibm/ds8k_certs/<host>.pem file.
Verify that the <host fqdn> is the same as configured in san_ip.

If the certificate subject and issuer are different, the certificate is signed by a CA, as illustrated below:

---
Certificate chain
0 s:/C=US/ST=New York/L=Armonk/O=IBM/OU=EI/CN=www.ibm.com
i:/C=US/O=GeoTrust Inc./CN=GeoTrust SSL CA - G3
1 s:/C=US/O=GeoTrust Inc./CN=GeoTrust SSL CA - G3
i:/C=US/O=GeoTrust Inc./CN=GeoTrust Global CA
---

Add a public certificate to trusted CA certificate store to complete the chain of trust, as explained below.
Verify trusted communication link, as explained below.

Adding a public certificate to trusted CA certificate store

Add the CA public certificate to the trusted CA certificates store on the Cinder node, according to procedures for the operating system in use.

For RHEL 7.x or CentOS 7.x, place the certificate to be trusted (in PEM format) into the /etc/pki/ca-trust/source/anchors/ directory. Then, run the sudo update-ca-trust command.
For Ubuntu 16.04, place the certificate to be trusted (in PEM format) into the /usr/local/share/ca-certificates/ directory. Rename the file, using the *.crt extension. Then, run the sudo update-ca-certificates command.
For Python requests library with certifi, run the cat ca_public_certificate.pem command to append the certificate to the location of the certifi trust store file. For example:
```
cat ca_public_certificate.pem >> /usr/local/lib/python2.7/
dist-packages/certifi/cacert.pem.
```

Verifying trusted communication link

Verify the chain of trust has been established successfully.

Obtain the location of the Python library requests trust store, according to the installation type.

RHEL 7.x or CentOS 7.x:

# python
    Python 2.7.5 (default, Oct 11 2015, 17:47:16)
    [GCC 4.8.3 20140911 (Red Hat 4.8.3-9)] on linux2
    Type "help", "copyright", "credits" or "license" for
    more information.
    >>> import requests
    >>> print requests.certs.where()
    /etc/pki/tls/certs/ca-bundle.crt

Ubuntu 16.04:

# python
    Python 2.7.6 (default, Jun 22 2015, 17:58:13)
    [GCC 4.8.2] on linux2
    Type "help", "copyright", "credits" or "license" for
    more information.
    >>> import requests
    >>> print requests.certs.where()
    /etc/ssl/certs/ca-certificates.crt

Python requests library with certifi:

# python
    Python 2.7.6 (default, Jun 22 2015, 17:58:13)
    [GCC 4.8.2] on linux2 Type "help", "copyright", "credits"
    or "license" for more information.
    >>> import requests
    >>> print requests.certs.where() /usr/local/lib/python2.7/
    dist-packages/certifi/cacert.pem

Run the openssl s_client -CAfile <location> -connect <host fqdn>:8452 </dev/null command. The following return codes indicate a successful or failed attempt in establishing a trusted communication link.

Verify return code: 0 (ok): success.
Verify return code: 21 (unable to verify the first certificate), or any other non-zero value: failure.

Troubleshooting

Refer to this information to troubleshoot technical problems that you might encounter when using the IBM Storage Driver for OpenStack.

Checking the Cinder log files

The Cinder log files record operation information that might be useful for troubleshooting.

To achieve optimal and clear logging of events, activate the verbose logging level in the cinder.conf file, located in the /etc/cinder folder. Add the following line in the file, save the file, and then restart the cinder-volume service:

verbose = True
debug = True

To turn off the verbose logging level, change True to False, save the file, and then restart the cinder-volume service.

Check the log files on a periodic basis to ensure that the IBM Storage Driver is functioning properly. To check the log file on a Cinder node, go to the /var/log/cinder folder and open the activity log file named cinder-volume.log or volume.log. The IBM Storage Driver writes to this log file using the [IBM DS8K STORAGE] or [IBM XIV STORAGE] prefix (depending on the relevant storage system) for each event that it records in the file.

Best practices

This section contains the general guidance and best practices.

Working with multi-tenancy (Spectrum Accelerate Family)

The XIV storage systems, running microcode version 11.5 or later, Spectrum Accelerate and FlashSystem A9000/A9000R can employ multi-tenancy.

In order to use multi-tenancy with the IBM Storage Driver for OpenStack:

For each storage system, verify that all predefined storage pools are in the same domain or, that all are not in a domain.
Use either storage administrator or domain administrator user's credentials, as long as the credentials grant a full access to the relevant pool.
If the user is a domain administrator, the storage system domain access policy can be CLOSED (domain_policy: access=CLOSED). Otherwise, verify that the storage system domain access policy is OPEN (domain_policy: access=OPEN).
If the user is not a domain administrator, the host management policy of the storage system domain can be BASIC (domain_policy: host_management=BASIC). Otherwise, verify that the storage system domain host management policy is EXTENDED (domain_policy: host_management=EXTENDED).

Working with IBM Real-time Compression™ (Spectrum Accelerate Family)

XIV storage systems running microcode version 11.6 or later, Spectrum Accelerate and FlashSystem A9000/A9000R can employ IBM Real-time Compression™.

Follow these guidelines when working with compressed storage resources using the IBM Storage Driver for OpenStack:

Compression mode cannot be changed for storage volumes, using the IBM Storage Driver for OpenStack. The volumes are created according to the default compression mode of the pool. For example, any volume created in a compressed pool will be compressed as well.
The minimum size for a compressed storage volume is 87 GB.

Working with QoS (Spectrum Accelerate Family)

The IBM Storage Driver for OpenStack provides QoS per volume for IBM FlashSystem A9000/A9000R storage systems, running microcode version of 12.0 or later. With QoS classes, the user can control the maximum bandwidth and I/O operations for each volume. For detailed instructions on QoS configuration, refer to the user documentation of the relevant storage system on IBM Knowledge Center.

QoS class types:

Shared (default). Limits the combined rates of all of the volumes in the same QoS class. The maximum rate is the sum of the combined rate for each volume. For example, two volumes under a QoS class of maximum 100 Gbps are allocated a combined maximum bandwidth rate of 100 Gbps.
Independent. Sets the maximum rate separately for each volume in the QoS class. For example, for two volumes under a QoS class of maximum 100 Gbps, each volume is limited to a rate of 100 Gbps. Thus, the combined maximum bandwidth rate is up to 200 Gbps.

To define a QoS class:

Create the QoS class:

cinder qos-create <class_name> <class_specs: bw=#, iops=#>

Create a type:

cinder type-create type_<qos_class_name>

Associate the QoS class with the type:

cinder qos-associate <qos uuid> <type uuid>

Announce that the type is supporting QoS:

cinder type-key <type_name or UUID> set QoS_support=True

Create a volume:

cinder create 1 --volume-type <type_name>

Configuring volume replication (DS8000 Family)

Volume replication is required for disaster recovery and high-availability applications running on top of OpenStack-based clouds. The IBM Storage Driver for OpenStack supports synchronous (Metro Mirror) volume replication for DS8000 storage systems.

Verify that:
- Master and remote storage pools exist on DS8000 systems.
- Reliable communication link is established between the primary and secondary sites, including physical connection and PPRC path.
- Metro Mirror replication is enabled on DS8000 storage systems.

Perform the following procedure, replacing the values in the example with your own:

enabled_backends = ibm_ds8k_1, ibm_ds8k_2
[ibm_ds8k_1]
proxy = cinder.volume.drivers.ds8k_proxy.DS8KProxy
volume_backend_name = ibm_ds8k_1
san_clustername = P2,P3
san_password = actual_password
san_login = actual_username
san_ip = host_fqdn
volume_driver = cinder.volume.drivers.ibm.ibm_storage.IBMStorageDriver
chap = disabled
connection_type = fibre_channel
replication_device = connection_type: fibre_channel,
backend_id: bar, san_ip: host_fqdn,
san_login: actual_username, san_password: actual_password,
san_clustername: P4, port_pairs: I0236-I0306; I0237-I0307

[ibm_ds8k_2]
proxy = cinder.volume.drivers.ibm.ds8k_proxy.DS8KProxy
volume_backend_name = ibm_ds8k_2
san_clustername = P4,P5
san_password = actual_password
san_login = actual_username
san_ip = 10.0.0.1
volume_driver = cinder.volume.drivers.ibm.ibm_storage.IBMStorageDriver
chap = disabled
connection_type = fibre_channel

Configuring groups

The IBM Storage Driver for OpenStack supports volume grouping. These groups can be assigned a group type, and used for replication and group snapshotting.

Replication groups

For better control over replication granularity, the user can employ volume grouping. This enables volume group replication and failover without affecting the entire backend. The user can choose between a generic group replication and consistency group (CG) replication. For consistency group replication, the driver utilizes the storage capabilities to handle CGs and replicate them to a remote site. On the other hand, in generic group replication, the driver replicates each volume individually. In addition, the user can select the replication type.

To configure group replication:

Create sync replicated consistency-group.

Create a volume type for replication.

#cinder type-create rep-vol-1

Create a volume type for replication.

#cinder type-key rep-vol-1
set replication_type='<is> sync'
replication_enabled='<is> True'

Create a group type.

#cinder group-type-create rep-gr-1

Configure the group type.

#cinder group-type-key rep-gr-1 set group_replication_enabled='<is> True' replication_type='<is> sync'

Create a replicated group, using existing group type and volume type.

#cinder group-create rep-gr-1 rep-vol-1 --name replicated-gr-1

Create a volume and add it to the group.
- Create a replicated volume.
```
#cinder create --name vol-1 --volume-type rep-vol-1 1
```
- Add the volume to the group.
```
#cinder group-update --add-volumes 91492ed9-c3cf-4732-a525-60e146510b90 replicated-gr-1
```
Note

You can also create the volume directly into the group by using the --group-id parameter, followed by ID of a group that the new volume belongs to. This function is supported by API version 3.13 and later.

Enable replication.

#cinder group-enable-replication replicated-gr-1

Disable replication.

#cinder group-disable-replication replicated-gr-1

Fail over the replicated group.

#cinder group-failover-replication replicated-gr-1

Consistency groups

Consistency groups are mostly the same as replication groups, but with additional support of group snapshots (consistent_group_snapshot_enabled parameter). See configuration example below.

#cinder group-type-create cg1
#cinder group-type-show cg1
#cinder group-type-key cg1 set consistent_group_snapshot_enabled="<is> True"
#cinder group-create --name cg1 IBM-DS8K_ibm.com_P0_P1_fibre_channel_not_thin,
IBM-DS8K_ibm.com_P0_P1_fibre_channel_thin,
IBM-DS8K_ibm.com_P0_P1_fibre_channel_not_thin_replica,
IBM-DS8K_ibm.com_P0_P1_fibre_channel_thin_replica

Using volume types for volume allocation control (DS8000 Family)

For better controls over volume placement granularity, you can use volume types. This enables volumes to be created on specific LSSes or pools. You can combine both types.

Storage pool

#cinder type-key pool-1_2 set drivers:storage_pool_ids='P1,P2'

LSS

#cinder type-key lss80_81 set drivers:storage_lss_ids='80,81'

29 KiB Raw Blame History