HP SAN Virtualization Services Platform 3.0.5 Release Notes (5697-1031, June 2011)
Table Of Contents
- HP StorageWorks SAN Virtualization Services Platform 3.0.5 Release Notes
- Description
- Versions
- Product models
- Fixes
- Prerequisites
- Installation instructions
- Host operating systems and multipath drivers
- Devices supported
- SVSP-supported storage arrays
- Capabilities and limitations
- SVSP system capabilities and limitations
- SVSP front-end capabilities and limitations
- SVSP back-end capabilities and limitations
- Configuration best practices
- Operational best practices
- Issues and workarounds
- Upgrade issues
- Upgrading with large back-end disks
- Upgrade order
- Replacing a DPM in a DPM group when upgrading from SVSP 2.x to SVSP 3.0.x
- Operating system issues
- Decreased DPM performance in VMware environment
- The Windows Server 2008 UAC prevents applications requiring admin privileges from working
- Windows offers to format newly discovered disks resulting in data loss
- Miscellaneous Windows issues
- OpenVMS support without async mirrors
- SMI-S provider can delete some setup virtual disks
- Servers running HP-UX 11.23 disconnect from SAN
- Setting the cluster bit for Serviceguard in Linux or HP-UX
- Capacity limits with thin provisioned volumes
- VSM issues
- Thin virtual disks must have read/write permission and cluster flag
- VSM GUI indicates passive VSM not ready for failover
- Disabled Delete menu when virtual disk is member of synchronous mirror group
- SVSP web server hang makes GUIs inaccessible
- WSRM configuration step needed on VSM v2 server
- Imported LUNs must be aligned to a 1 MB capacity
- Verify free space before concatenating thin virtual disk on thick virtual disk
- Manual failover of loaded system with VSM GUI causes a non-expected system state
- VSM GUI unable to remove all host presentations in one operation
- Threshold limit message only sent once
- Restoring a VSM setup from a backup cannot have PiTs
- Error 1618 during VSM installation
- Limit of 63 PiTs on thin virtual disk
- Front-end hosts not able to work with objects
- Export/import query does not work
- Cannot have same first 29 characters in sync mirror task names
- Cannot expand VDG without same DPM as the members
- VSM does not enforce 512 data movers
- Issue when exceeding 180 async mirror tasks
- MSA capacity can be reported in base 2 or base 10
- Error with Continuous Access license capacity when replicated differently
- HP XP array volumes show as failed after creation of storage pool
- Long boot time with large configurations
- Cannot write to file error when using an MSA22xx array
- ForceActive with VSM server in unknown state
- Deletion of last DPM in a DPM group
- Changing virtual disk or group DPM selection from two to one DPM is not allowed
- Windows fault when array unresponsive
- VSM must be enabled for "Reboot on fatal error"
- VSM server connection to the SAN
- New front-end LUNs require rescan
- New back-end LUNs require rescan
- LUNs must have a preferred path/controller
- Import some HP and IBM back-end LUs only to designated pools
- Asynchronous mirrors and jobs do not fail when the source site is unavailable
- Accessing VSM may require Java Web Services cleanup
- IPv6 support limitations
- PiT deletions after reaching threshold Guard level
- Read-only permissions are not supported in the VSM GUI
- Potential for I/O loss with synchronous mirrors due to active DPM failure
- DPM issues
- DPM ports stay present during a DPM reboot
- Failback operation does not work after DPM replacement
- Physically disconnecting a DPM port results in a high rate of invalid transmitted word port errors
- Changing back-end virtual disks to be write protected is not supported
- Enabling Ethernet port without a cable results in diagnostic message on console
- A DPM upgrade reverts to SNMP host name setup
- Change active switch command does not work with Sun Solaris hosts
- Rescan is required after deleting or unpresenting virtual disks
- Period of high back-end load after a DPM reboot when synchronous mirror virtual disks are used
- I/O errors/timeouts when performing failover on large numbers of synchronous mirror virtual disks
- DPM reboots when a host has two different host types or personalities on two different SVSP domains
- DPM ports connected to a 4 Gb Fibre Channel switch may sometimes auto-negotiate to 2 Gb speed
- DPM NTP configuration does not work with Windows 2000 domain controllers
- Initial GUI port statistics values are zero for an extended time
- Running show debug dump command does not work as expected
- Order of magnitude of Tx/Rx Kbytes not correct with show ports stats command
- The ‘%’ option of the show log command moves to the end of the file
- Additional DPM limitations
- HP Command View SVSP issues
- HP Command View SVSP has reached end-of-life
- VSM CLI issues
- The SelectVSMDomain command fails with certain operating systems
- Some VSM CLI commands cause the CLI application to stop
- No error message with PurgeVirtualDisks command
- Exception error using Solaris with no VSM CLI virtual disk
- Host not checked when creating snapclone group
- Error received with GetDpmInfo command when using –SyncGroup argument
- Incorrect error message displayed with CreateSnapshotWithLun command
- Incorrect error message when ExpandVirtualDisk command is missing argument
- Incorrect error message when creating 65th PiT on a virtual disk
- The ListSnapshots command fails to return data with too many PiTs
- The DeleteRemoteAsyncMirrorGroup command does not work in HP-UX
- The BreakAllRemoteAsyncMirrorTasks command does not work in HP-UX
- The ? wildcard character in HP-UX is not reliable
- The ResumeLocalAsyncMirrorTask command does not work in HP-UX the first time
- The ResumeLocalAsyncMirrorTask command gives error in HP-UX if already resumed
- Incorrect error message when ManageSnapcloneGroup command is missing the Opcode argument in HP-UX
- The CreateSnapcloneGroupEx command fails with HP-UX if the group is still building
- The CreateSnapshotWithLun command fails to get last PiT instance for Pit26 with HP-UX
- Do not use DeleteVirtualDisk command with LastInstance argument in HP-UX
- Error with DeleteVirtualDisk command when virtual disk LUN number is out of range
- The CreateAsyncMirrorGroupEx command cannot be run in Linux
- VSS hardware provider issues
- Documentation issues
- Obtaining product documentation
• Windows only recognizes 254 LUNs with the HP DSM driver though the Windows operating
system supports 255 LUNs (maximum).
• Automatic failback of Microsoft Cluster Service resource groups does not occur when both the
HBA cables of one node are pulled out and reinserted.
OpenVMS support without async mirrors
OpenVMS 8.3 is supported with the exception of asynchronous mirrors.
SMI-S provider can delete some setup virtual disks
SVSP normally uses three setup virtual disks that are a part of a synchronous mirror group. It is possible
while using the SMI-S provider to delete two of these setup virtual disks, but not the third. This can
cause system availability issues, so ensure that care is taken to not delete any setup virtual disks.
Servers running HP-UX 11.23 disconnect from SAN
Servers running HP-UX 11.23 have been seen to disconnect from the SAN if no logical units are
presented to them, and SVSP has been added to the configuration. The workaround is to temporarily
present a unit directly from the storage arrays to the host and restart or perform an ioscan. This causes
the server to remain on the SAN long enough to be discovered by SVSP.
Setting the cluster bit for Serviceguard in Linux or HP-UX
If virtual disks are presented through Data Path Modules with Linux or HP-UX, do not set the cluster bit,
as this will force the DPMs to abort pending writes after a period of time.
Capacity limits with thin provisioned volumes
When you run out of free space in a storage pool, a write to a thin provisioned virtual disk will create
an error to the I/O, but information about that error may not convey the idea that thin volume space
has run out because the SCSI standards have not taken this into account yet. You should receive a
message or alert from the VSM indicating that space for a thin volume has run out, because the VSM
understands thin volumes. This can be prevented by properly setting and actively monitoring your pool
protection thresholds, adding capacity, or adding thin provisioning licenses.
VSM issues
Thin virtual disks must have read/write permission and cluster flag
Do not create thin virtual disks with:
• Only read-only permissions (in other words, without any read-write permissions)
• Read-write permissions to both VSMs without the cluster flag.
If either operation is performed, the operation is blocked, but the virtual disk is created in a Delete
state. The thin virtual disk is deleted only upon restart or failover of the active VSM.
VSM GUI indicates passive VSM not ready for failover
The VSM GUI may erroneously indicate that the passive VSM server is not ready for a VSM failover
operation (for example, the passive VSM does not detect all the paths seen by the active VSM). This
is usually incorrect, and performing a VSM failover completes successfully.
12 Issues and workarounds