V8.2.1.x Configuration Limits and Restrictions for IBM Storwize V7000 and V7000F

1. Child pools are not supported in a DRP;
2. VVOL is not supported in a DRP (because child pool is not supported);
3. A volume in a DRP cannot be shrunk;
4. No volume move between I/O groups if volume in a DRP (use FlashCopy or Metro Mirror/Global Mirror instead);
5. No split of a volume mirror to copy in a different I/O group;
6. Real/used/free/free/tier capacity are not reported per volume - only per pool.

Note: These restrictions are applicable to all the versions of IBM Spectrum Virtualize v8.1.2 and later

REST API

Customers using the REST API to list more than 2000 objects may experience a loss of service, from the API, as it restarts due to memory constraints.

It is not possible to access the REST API using a cluster's IPv6 address.

NVMe over Fibre Chanel

Hosts using the NVMe protocol cannot be mapped to HyperSwap or stretched volumes.

Volumes accessed by hosts using the NVMe protocol cannot be configured with multiple access I/O groups due to a limitation of the NVMe protocol.

RAID and Distributed RAID

Storwize V7000 Gen 3 systems support DRAID5, DRAID6 and RAID10. RAID5 and RAID6 are not supported.

DRAID Strip Size

For candidate drives, with a capacity greater than 4TB, a strip size of 128 cannot be specified for either RAID-5 or RAID-6 DRAID arrays. For these drives a strip size of 256 should be used.

Non-Disruptive Volume Move (NDVM)
The following Fibre Channel attached host types are supported for non-disruptively moving a volume between I/O groups:

Note: For all other host types, I/O should be quiesced when moving a volume.

When moving a volume that is mapped to a host cluster then you must rescan disk paths on all host cluster nodes to ensure the new paths have been detected before removing access from the original I/O group.

NPIV ( N_Port ID Virtualization )

SAN Volume Controller and Storwize Version 7.7 introduced support for NPIV ( N_Port ID Virtualization ) for Fibre Channel fabric attachment. FCoE is not supported with NPIV. The following recommendations and restrictions should be followed when implementing the NPIV feature.

Operating systems not currently supported for use with NPIV:

• RHEL6 and earlier on IBM Power
• HPUX 11iV2
• Veritas DMP multipathing on Windows with RAID-5 volumes in VxVM

General requirements

Required SDD versions for IBM AIX and Microsoft Windows Environments:

1. IBM AIX Operating Systems require a minimum SDDPCM version of 2.6.8.0 or AIXPCM;
2. Microsoft Windows requires a minimum SDDDSM version 2.4.7.0. The latest recommended level which resolves issues listed below is 2.4.7.1.

Path Optimization

User intervention may be required when changing NPIV states from "Transitional" to "Disabled". All Paths to a LUN with SDDDSM or SDDPCM may remain "Non-Optimized" when NPIV is "Disabled" from "Transitional" state.

IBM AIX
For SDDPCM:
    Run "pcmpath chgprefercntl device <device number>/<device number range>" on AIX. This will restore both Optimized and Non-Optimized paths for all the LUNs correctly.

Windows 2008 and 2012
For SDDDSM:
    Run "datapath rescanhw" on Windows. This will restore both Optimized and Non-Optimized paths for all the LUN's correctly.
This issue is resolved with SDDDSM version 2.4.7.1

Windows 2008 and 2012 Non-Preferred Paths with SDDDSM
When NPIV enters into Transitional state from Disabled, with all the SDDDSM paths in Non-Preferred state, the paths to the Virtual ports also become Non-Preferred. This path configuration might cause IO failures as soon as NPIV moves into "Enabled" state.
As a work around user should configure "at least one preferred path" to each LUN, when in NPIV "Disabled" state.
This issue is resolved with SDDDSM version 2.4.7.1

Solaris
Emulex HBA Settings:
1. When implementing NPIV on Solaris 11 the default disk IO timeout needs to be changed to 120s by adding "set sd:sd_io_time=120" in /etc/system file, A system reboot is required for the change to be implemented.
2. When ports on host HBA are connected to 16GB SAN, NPIV is not supported.

Other Operating Systems
Other operating Systems may also experience the same issue when changing the NPIV state from "Transitional" to "Disabled" in which case the operating system specific rescan command should be used.

Fabric Attachment
NPIV mode on SVC or Storwize is only supported when used with Brocade or Cisco fibre channel SAN switches which are NPIV capable.

Nodes in an IO group cannot be replaced by nodes with less memory when compressed volumes are present

If a customer must migrate from 64GB to 32GB memory node canisters in an IO group, they will have to remove all compressed volume copies in that IO group. This restriction applies to 7.7.0.0 and newer software.

A customer must not:

1. Create an IO group with node canisters with 64GB of memory.
2. Create compressed volumes in that IO group.
3. Delete both node canisters from the system with CLI or GUI.
4. Install new node canisters with 32GB of memory and add them to the configuration in the original IO group with CLI or GUI.

HyperSwap

When using the HyperSwap function please configure your host multipath driver to use an ALUA-based path policy.

Due to the requirement for multiple access IO groups, SAS attached host types are not supported with HyperSwap volumes.

A volume configured with multiple access I/O groups, on a system in the storage layer, cannot be virtualized by a system in the replication layer.  This restriction prevents a HyperSwap volume on one system being virtualized by another.

AIX Live Partition Mobility (LPM)
AIX LPM is supported with the HyperSwap function and AIX 7.x 

Clustered Systems

A Storwize V7000 system requires native Fibre Channel SAN or alternatively 8Gbps/16Gbps Direct Attach Fibre Channel connectivity for communication between all nodes in the local cluster. Fibre Channel over Ethernet ( FCoE ) connectivity for communication between all nodes in the local cluster is also supported. Clustering can also be accomplished with 25Gbps Ethernet, for standard topologies. For HyperSwap topologies a SCORE request will be required. Please contact your IBM representative to raise a SCORE request.

Partnerships between systems for Metro Mirror or Global Mirror replication can be used with both Fibre Channel, Native Ethernet connectivity and FCoE connectivity, however direct FCoE links are only supported to a maximum of 300 metres. Distances greater than 300 metres are only supported when using an FCIP link or Fibre Channel between source and target.

Direct Attachment

SAN boot is not supported on host Fibre Channel direct attach systems running v8.2.1.0 - v8.2.1.2. This configuration is supported with v8.2.1.3 or later.

Cisco Nexus

The minimum level of Cisco Nexus firmware supported for FCoE with the IBM Storwize V7000 Gen2 / Gen2+ is 5.2(1)N1(2a).

16Gbps Fibre Channel Canister Connection

Please visit the IBM System Storage Inter-operation Center (SSIC) for Fibre Channel configurations supported with 16Gbps node hardware.

Note 16Gbps Node hardware is supported when connected to Brocade and Cisco 8Gbps or 16Gbps fabrics only.

Direct connections to 2Gbps or 4Gbps SAN or direct host attachment to 2Gbps or 4Gbps ports is not supported.

Other configured switches which are not directly connected to the 16Gbps Node hardware can be any supported fabric switch as currently listed in SSIC.

25Gbps Ethernet Canister Connection

Two optional 2-port 25Gbps Ethernet adapter is supported in each node canister for iSCSI communication with iSCSI capable Ethernet ports in hosts via Ethernet switches. These 2-port 25Gbps Ethernet adapters do not support FCoE.

There are two types of 25Gbps Ethernet adapter feature supported:

1. RDMA over Converged Ethernet (RoCE)
2. Internet Wide-area RDMA Protocol (iWARP)

Either will work for standard iSCSI communications, i.e. not using Remote Direct Memory Access (RDMA). A future software release will add (RDMA) links using new protocols that support RDMA such as NVMe over Ethernet.

When use of RDMA with a 25Gbps Ethernet adapter becomes possible then RDMA links will only work between RoCE ports or between iWARP ports (i.e. from a RoCE node canister port to a RoCE port on a host or from an iWARP node canister port to an iWARP port on a host).

The 25Gbps adapters come with SFP28 fitted, which can be used to connect to switches using OM3 optical cables.

For Ethernet switches and adapters supported in hosts visit the SSIC.

This is an example of a RoCE adapter for use in a host.
https://docs.nvidia.com/networking/display/cx4lxen

This is an example of a iWARP adapter for use in a host.
https://www.chelsio.com/nic/unified-wire-adapters/t6225-cr/

Customers who want to connect a 10Gb switch to a 25Gb HBA should be aware that this is only supported via a SCORE request. Please contact your IBM representative to raise a SCORE request.

IP Partnership

IP partnerships are supported on any of the available ethernet ports. Using an Ethernet switch to convert a 25Gb to a 1Gb IP partnership, or a 10Gb to a 1Gb IP partnership is not supported. Therefore the IP infrastructure on both partnership sites must match. Bandwidth limiting on IP partnerships between both sites is supported.

Fabric Limitations

Only one FCF ( Fibre Channel Forwarder ) switch per fabric is supported.

VMware vSphere Virtual Volumes (VVols)

The maximum number of Virtual Machines on a single VMware ESXi host in a Storwize / VVol storage configuration is limited to 680.

The use of VMware vSphere Virtual Volumes (VVols) on a system that is configured for HyperSwap is not currently supported with SVC/Storwize.

Host Limitations

Cisco UCS
UCS firmware code level: 4.0(4)
UCS Platforms M3, M4 and M5

Protocol: FC (EHM)

• UEFI SAN boot Is not supported with VIC13XX adapters
• Path recovery via unified FC uplink port fails when the attached switch rebooted

Protocol: ISCSI

• Connecting the Storwize Storage Array (either of the 25Gb HICs) to a UCS system must be done via breakout cables.

iSER
Operating systems not currently supported for use with iSER:

• VMware ESXi 6.7 using Mellanox ConnectX-4 Lx EN
• Windows 2012 R2 using Mellanox ConnectX-4 Lx EN
• Windows 2016 using Mellanox ConnectX-4 Lx EN

FCoE
Operating systems not currently supported for use with FCoE:

• RedHat 6.x
• VMware 6.0
• Windows 2012 Hyper-V Cluster

Microsoft Offload Data Transfer ( ODX ) and SDDDSM Requirements
Storwize V7000 version 7.5.0 introduced support for Microsoft ODX. In order to utilise this function all windows hosts accessing Storwize V7000 are required to be at a minimum SDDDSM version of 2.4.5.0. Earlier versions of SDDDSM are not supported when the ODX function is activated.

Windows NTP server 
The Linux NTP client used by Storwize V7000 may not always function correctly with Windows W32Time NTP Server

Oracle

Restriction 1:
Oracle ASM disk groups may dismount with the following error: 

"Waited 15 secs for write IO to PST"

Recommendation

Increase the asm_hbeatiowait to 120 seconds to prevent this issue occurring.

Applies to Oracle Database - Enterprise Edition - Version 11.2.0.3 to 12.1.0.1 [Release 11.2 to 12.1] on any platform

Priority Flow Control for iSCSI

Priority Flow Control for iSCSI is supported on Brocade VDX 10-gigabit Ethernet switches only.

Maximum Configurations

Configuration limits for Storwize V7000:

Extents

The following table compares the maximum volume, MDisk and system capacity for each extent size.

* The total capacity values assumes that all of the storage pools in the system use the same extent size.
** Please see the  following Flash