HP-UX vPars and Integrity VM V6.3 Release Notes (762790-001, July 2014) (Edition: 1.6)
Table Of Contents
- HP-UX vPars and Integrity VM V6.3 Release Notes
- Contents
- HP secure development lifecycle
- 1 Introduction
- 2 Installing or upgrading to HP-UX vPars and Integrity VM V6.3
- 3 New functionality and changes from earlier versions
- 3.1 New features and enhancements
- 3.1.1 Enhanced capability for emulated platform NVRAM (Non Volatile RAM)
- 3.1.2 Increased resources for Integrity VM guests
- 3.1.3 Dynamic addition of I/O devices
- 3.1.4 PCI OLR support on Superdome 2 VSPs
- 3.1.5 AVIO Networking improvements
- 3.1.6 AVIO Storage improvements
- 3.1.7 Greater flexibility for online VM migration
- 3.1.8 Improvements to Virtual Server Management
- 3.1.9 Improvements to VSP resource management
- 3.2 Changes from previous versions
- 3.1 New features and enhancements
- 4 Known problems, limitations, and workarounds
- 4.1 CPU/vCPU
- 4.2 Memory
- 4.3 Networking
- 4.3.1 Cannot remove a VLAN-based vNIC if the VLAN has been removed
- 4.3.2 hpvmhwmgmt might add ports in link aggregates into the DIO pool
- 4.3.3 DIO limitations
- 4.3.4 Known issues or limitations with DIO support for 10GigEthr-02 (iexgbe)
- 4.3.5 DIO-capable functions might become inconsistent with information in vPar or VM device database
- 4.3.6 When DIO device is assigned or removed from the DIO pool, error messages appear multiple times
- 4.4 Storage
- 4.4.1 Presenting a Logical Volume created on iSCSI devices as AVIO backing store to a guest not supported
- 4.4.2 Size change operations on a SLVM volume based backing store do not get reflected in the vPar or VM
- 4.4.3 The hpvmdevinfo command may not list the correct host to guest device mapping for legacy AVIO backing stores
- 4.4.4 Probe of NPIV HBAs for Fibre Channel targets may timeout
- 4.4.5 NPIV LUNs not shown by default invocation of ioscan
- 4.4.6 The interrupt balancing daemon must not be enabled in vPars and Integrity VM guests
- 4.4.7 Online addition of DMP device as backing store is not supported
- 4.5 VM <—> vPar conversion
- 4.6 Migration, Suspend, and Resume operations on Integrity VM guests
- 4.6.1 Use of -F with hpvmmigrate on a suspended VM can cause VM to be not runnable on both source and target
- 4.6.2 Copy of a vPar or VM might be left in runnable state if migration fails
- 4.6.3 Interrupt migration of vNICs during Online guest migration can result in network disconnectivity
- 4.6.4 Physical NIC link state change during hpvmsuspend to hpvmresume may result vNIC in down state
- 4.6.5 Offline migration of a guest- with multiple DIO resources might succeed with errors if the DIO devices are added under the same label
- 4.7 User interface—CLI
- 4.8 Known system crashes, panics, hangs and MCAs
- 5 HP-UX vPars and Integrity VM support policy
- 5.1 Support duration
- 5.2 VSP firmware requirements
- 5.3 VSP server and OS support
- 5.4 HP-UX version support for vPar and Integrity VM guests
- 5.5 Storage device support for vPar and Integrity VM guests
- 5.6 Network device support for vPar and Integrity VM guests
- 5.7 Support for migration of vPars and Integrity VMs
- 6 Support and other resources
- 7 Documentation feedback
esvroot_node_cb + 0x1e0
wsio_cdio_cb + 0xc0
io_node_callback + 0x220
gio_scan_all + 0x870
ioconfig_dispatcher + 0x30
kthread_daemon_startup + 0x90
The problem happens only on servers based on the Intel® Itanium® 9500 processor series, when
the length of the backing store name for the following type of devices is between 14 and 17
characters:
• LVM: /dev/vg_name/rlvol_name
• VxVM: /dev/vx/rdisk/dg_name/v_name
• FILE backing stores: /pathname/file
The issue is fixed in patch PHKL_42989 (esctl cumulative patch) which is part of HP-UX 11i v3
September 2012 (AR1209) release. HP recommends to use AR1209 or newer media when installing
VM guests or vPars on Itanium 9500 systems.
Workaround
The vPar or VM guest can be installed onto a disk based backing store or a backing store with a
longer name, and additional devices with affected name lengths can be added after the
PHKL_42989 (esctl cumulative patch) patch is installed onto the guest OS.
4.8.2 VSP might crash when booting guest with more than 60 vNICs
The VSP might crash when a VM with more than 60 vNICs is started or shortly after the boot.
To avoid this potential crash, limit the number of vNICs on a guest to 60.
4.8.3 The 11i v2 guest might temporarily hang under certain conditions
When the HP-UX 11i v2 VM guests are configured to run with 16 vCPUs on a system with two or
more other VM guests, the 11i v2 guest might experience a temporary hang due to heavy CPU
and I/O intensive load conditions. A message “watchdog timeout” is seen on the guest console.
Workaround
Configure the 11i v2 guests with a maximum of 8 vCPUs.
4.8.4 MCA may occur while stopping a vPar or a VM guest with DIO interface
While adding a DLA (Device Level Assignment) device to the DIO pool using the hpvmwhmgmt
command, HPVM ensures that all the functions (ports) of DLA are moved to the pool. But, there is
a scenario which could lead to breaking this rule and could result in an MCA of the VSP.
When using VC (Virtual Connect), a DLA device which is not yet configured in VC has only two
ports visible from the HP-UX perspective (and both these ports are moved to DIO pool if any of
them is added using the hpvmhwmgmt command). When such a device is later configured in the
Virtual Connect Manager, the number of functions seen from HP-UX perspective changes to eight.
The original two ports are still in DIO pool owned by hpvmdio driver, but the new six ports get
claimed by their native driver (iexgbe).
While stopping a guest which uses such DIO function, in certain scenarios a Global MCA may
happen, bringing the VSP down.
To prevent this, HP recommends to configure VC before adding device to DIO pool or remove the
functions from the DIO pool and add them again after the VC configuration. For more information
about removing or adding cards from or to DIO pool, see the HP-UX vPars and Integrity VM
Administrator Guide or hpvmhwmgmt(1M) man page.
Only DLA devices (claimed by iexgbe driver) are potentially affected by this problem.
4.8 Known system crashes, panics, hangs and MCAs 27