HP 3PAR InForm OS 3.1.1 MU1 Release Notes

DescriptionItemBug ID
During a power failure or multi-node panic recovery, there were some cases
in which the system required user intervention in the form of issuing a
Recovery requires user
intervention
58852
setsysmgr force_iderecovery command. These cases were a) when
all but one node had saved data without proper handshake with the missing
node and b) when the system was in the process of creating a volume (base
or snapshot) when the save occurred. A fix has been implemented that
eliminates the need for user intervention; the system will now recover data
automatically.
For a virtual volume, if the metadata view of size is smaller than actual size,
the metadata checker will now fail the volume. Also while probing for holes
Virtual volume with metadata
issues does not fail properly
58886
in snapspace due to reclaim, if metadata view cannot account for all the
holes, the metadata checker will now fail the volume.
During a power failure or multi-node panic recovery, if the storage system
encountered problems paging in volume metadata, recovery could cause
Recovery requires user
intervention
58939
all the nodes to panic. When the storage system restarted, it required user
intervention in the form of a setsysmgr force_iderecovery command
on order to proceed with recovery. A fix has been implemented in which,
if the metadata page has failed due to disk unavailability, the system will
now automatically proceeds with recovery, and does not require user
intervention.
If the firmware on the 4-port 4GB Fibre Channel HBA experienced a fatal
error while I/O was ongoing to disk drives via the HBA, the firmware would
Disk drives unresponsive and
marked failed by storage
system
59114
not respond for several seconds while the transmit and receive lasers were
still active. This could lead to the disk drives becoming unresponsive for up
to a few minutes, resulting in the drives being marked failed by the storage
system. If disk cages were connected to multiple HBA ports, this could result
in multiple drives being marked failed leading to virtual volumes getting
marked unavailable. This has been resolved by having the firmware turn
off the transmit and receive lasers when it receives a fatal error, which
allows the drives to detect the link failure immediately and recover without
becoming unresponsive.
A system manager restart after an online upgrade had a small chance of
causing repeated system manager panics. This has been fixed.
System manager restart after
online upgrade causes panics
59274
The performance of SCSI-2 reservation handling has been improved to
allow more simultaneous incoming reservations.
Improved SCSI-2 reservation
handling performance
59549
After a master node restart or storage system restart, events from a long
time period prior could reappear in showeventlog. This has been fixed.
Old events appearing in
showeventlog
59672
After setting the port speed for Remote Copy links to 100Mbps from
1000Mbps the links failed due to negotiation issues. The links are recovered
Remote Copy links down after
changing port speed
60162
by setting auto negotiation for those links. The ability to set the speed to
rates less than 1000Mbps will be removed in a future release.
When system manager attempted to create a copy of itself to handle a
showspace request and that copy creation failed, the event that was
System manager copy
creation failure results in
repeated event filter panics.
60638
generated was not parsed correctly and resulted in repeated panics by the
event filter. Because this caused events to be queued in memory indefinitely,
an out of memory situation could result, especially in configurations using
remote copy. This has been fixed.
Each 4GB Fibre Channel port on the storage system has 511 Remote Port
Indicator (RPI) resources. One resource is used when an initiator logs in.
Remote Port Indicator not
released properly
61125
When the initiator logs out or gets disconnected, the RPI was not being
released. If the initiator logged back in with the same Fibre Channel D_ID,
the same RPI was being reused, but if it logged in with a different D_ID, the
original RPI was lost. After 511 such logins, the resource was exhausted
and all further logins were blocked. This was fixed by releasing the RPI
when an initiator logs out or is disconnected.
A miscompare was occurring between primary and secondary volumes in
the event of a node down. The miscompare happened when an I/O was
Miscompare between primary
and secondary volumes
61212
Modifications to the InForm OS 15