MU3 Release Notes

ManualsBrandsHP ManualsComputer AccessoriesHP 3PAR Inform S800/4x400GB Magazine LTU

DescriptionItemBug ID

The error message when an empty virutal volume set was exported was not

clear. Because exporting a virtual volume set isn't a logical action this action

is no longer allowed.

Empty virtual volume set

export error message

55218

Auto-choosing a LUN for a host that saw active exports using a host set

would sometimes incorrectly fail. The checks for host set exports are

improved to resolve this issue.

LUN auto-choose56082

When the firmware in the 2-port 4GB Fibre Channel HBA in the E and

F-series storage systems received incorrectly formatted Fibre Channel frames,

Defect in driver/firmware

results in port failure.

56195

the frame buffers were not released into the free buffer pool. When a large

number of these occurred, the port could run out of buffer credits and fail.

This was addressed by releasing these invalid frames correctly. This issue

was seen with a Cisco UCS server using the Palo interface card connected

to the storage system 4G port using a Brocade fabric when SAN booting

Windows 2008 R2. The Palo card driver/firmware had an issue causing

it to send incorrectly formatted Fibre Channel frames.

If a zerold was mapped to a volume offset 0, then the Inform Management

Console would display the incorrect RAID type. This issue is addressed.

RAID type displayed

incorrectly

56573

If an iSCSI port on a storage system received a non-login-related iSCSI PDU

from an initiator before the login command sequence was completed, the

InForm OS kernel panic due

to accessing an un-initialized

pointer.

56754

InForm OS kernel could panic due to accessing an uninitialized pointer.

This issue is addressed by processing the PDU correctly and sending a login

reject per the iSCSI protocol.

System manager does not start after a system or system manager restart

under the following scenario: running setvv -snp_cpg followed by

System manager is unable to

start up after panic,

repeatedly panics after the

initial panic.

57546

removing the CPG which was being used as a snap CPG before running

the command. Removal of the CPG could happen a few days later. Storage

systems with ESX clusters attached to them would have virtual host failures

because SCSI 2 reservations were not handled while system manager was

down. This issue is addressed.

The Link Power Down event is now properly cleared.The Link Power Down event

needs to be cleared to avoid

58215

a node being placed in a

degraded state after the node

rejoins.

An enhancement is implemented to correctly identify SCSI-3 reservation and

extend the check for duplicate events to include the node identifier and to

ensure that initiator list fields are populated correctly for duplicate events.

InForm OS kernel incorrectly

identifies SCSI-3 events

58269

The initial design of the PD scrubber was to scrub the entire system every

seven days. Systems have gotten much larger and delays were added to

Physical Disk (PD) scrub

performance issues

58586

the prevent the scrubber from impacting system performance. These two

changes have resulted in many systems being scrubbed at a much slower

rate (every 60 days or more). This opens a larger window for double disk

failures. The previous design kept one chunklet per loop active and would

issue one ioctl to kernel every scan step. The new design issues one

ioctl for one chunklet of all the PDs on the node. Inside the kernel scan

steps are scheduled based on a self-implemented timer, so that the scrubbing

rate can be controlled, VERIFY operations can be spread out evenly, and

multiple PD scans can be processed at the same time.

The InForm OS kernel could leak memory when processing some SCSI

commands, as memory buffers dynamically allocated for command

InForm OS kernel memory

leak

58703

processing are not freed on command completion. This leak could affect

certain commands received on LUNs to which no virtual volume is exported.

If a sufficiently large number of these commands were received, available

memory was exhausted and the node would panic. An enhancement is

implemented which ensures that all memory allocated for processing of

these commands is released on command completion.

14 HP 3PAR InForm OS 3.1.1 GA Release Notes