HP StorageWorks Fabric OS 5.2.3 Release Notes (AA-RWEYD-TE, November 2007)

In extremely rare occurrences, due to a race condition, a partial HA
updatearrivesbeforeafullupdatearrivesfromtheactiveCP.Asa
result, the standby CP goes into a continuous restart cycle while it
attempts to apply the partial HA updates.
Fixed per Fabric OS 5.2.2.
Port Mirrorin
g feature i s not supported on the HP StorageWorks 4/32B. Fixed per Fabric OS 5.2.2.
Third party system is sending PLOGI frames with its DID as a broadcast
format: 0x22ffffff. The 2/128 forwards the frame as a broadcast
frame, which causes other devices in the fabric to misbehave.
Fixed per Fabric OS 5.2.2.
CLI system verification incorrectly reports a false positive failure on the
FC4-16IP blade. If this diagnostic is run as part of a maintenance
action, the FC4-16IP blade could be incorrectly identified as faulty.
Fixed per Fabric OS 5.2.2.
Process logi
n drop due to devices in different edge fabrics with same
node WWN. This applies to 400 MP Router and 4/256 Director
with FCR-18i blade. Customer symptom could be that the host cannot
see the device, with the nam e server daemon queue full, etc. When
there are two devices with the same no de WWN connected to 2
edge fabrics, the FCR may not forward the NS request to the proper
domain, which results in the Name Server request timing out and
retrying every 2 seconds.
Fixed per Fabric OS 5.2.2.
Configdownload for Fabric Watch p arameters shows succ essful
message from Fabric Manager, but actually download fails for
non-disruptive configuration parameters such as Fabric Watch, SNMP,
Syslog and Time ser vice. Download configuration parameter from
Fabric Manager reports as successful, but n ew parameters for Fabric
Watch, SNMP, etc. are not accepted and the switch continues to
operate with the old configuration parameters.
Fixed per Fabric OS 5.2.2.
Traffic cannot be routed into backbone fabric from an EX_port on
400 MP Router with domain ID 7 or 12, and from an EX_port on
4/256 with FR4-18i blade with various fabric domain IDs causing
connectivity problem between two edge fabrics such as the host not
seeing a device, or traffic loss, etc. This is due to incorrectly filtering
out certain domain IDs.
Fixed per Fabric OS 5.2.2.
With
4Gbps platforms with FL ports connected, when the LIP is
rece
ived while the switch is no longer in the state of listening for L I P
(after loop initialization is finished and switch has sent out the CLS),
the LIP will cause the buffer to get stuck. This can also impact the
rec
eiving E_port due to buffer hold up on the FL port. The end result is
that the E_port or FL_port is faulted and host cannot see target.
Fix
ed per Fabric OS 5.2.2.
FICON CUP daemon crashes after switch comes up when it starts
processing the RNID command from FICON channels with Fabric OS
5.1 and later due to race condition in handling the RNID command.
Fixed per Fabric OS 5.2.2.
Hotcodeloadtimeon4/32Bcanexceedtheupperthresholdwhich
might cause a disruptive upgrade.
The fix in Fabric OS 5.2.2 enhances
thecodetoreducetheHCLtime,
placing it within the threshold.
The enc_out counter is increasing on unused ports on 2Gbps platforms
after HA reboot or firmware download operations.
W
ith Fabric OS 5.2.2, Enc Out
c
ounter is no longer incremented
if the por t is not active. Also
changes have been made to
reset these counters during HA
reboot/firmware download if the
port is not active.
SNMP traps are repeatedly sent from the switch every 2 seconds when
aportistakenoutofafaultystateimplicitlybyremovingthecable
or removing a module. Problem was caused by too m any port offline
SCNs being sent out during port fault and recover when there is laser
fault.
The fix in Fabric OS 5.2.2 sends
an offline SCN to the upper API
if and only if there is no light or
no module (Cover cable remove &
module remove cases only).
HP StorageWorks Fabric OS 5.2.3 release notes
21