HP-UX vPars and Integrity VM V6.3 Administrator Guide

Online migration of guests configured with NPIV HBAs fails; error messages indicate “data
put failure” and “invalid target”
Online migration of a guest configured with NPIV HBAs fails with the following message:
Target: dynamic IO data put failure - status 4 tag 0 length 0 depth 0
And, the target VSP syslog contains an error message from the host virtual storage driver similar
to the following:
HVSD: HPVM online migration error: invalid target id 0x207000c0ffda4ee1 under hba port 0x5001438002a30063 for
VM instance 1
This can be an indication that a target port that was visible from the source VSP is no longer visible
from the target VSP. This might occur if zoning configuration on the FC fabric shared by the source
and target VSP is incorrect. To ensure successful migration of guests with NPIV devices, HP
recommends that the SAN administrator uses WWN based zoning instead of Port based zoning.
This error can mean that a target port that was visible from the source VSP has failed and gone
offline. To be able to migrate the guest online, the failed or unavailable target must be cleaned
up from the guest prior to attempting a migration. This can be done by running a rmsf -H against
the target path in the guest.
For the rmsf command to clean up all the stale target information, the FC drivers or FCoC drivers
in the host must be March 2013 version or later. For more information about the list of dependencies,
see HP-UX vPars and Integrity VM Release Notes.
NPIV LUNs not shown by default invocation of ioscan
By default, the ioscan(1M) command displays only devices that use legacy style device file
format. NPIV LUNs use the agile device file format. The -N option to ioscan must be specified in
order to display NPIV LUNs.
A.2.3 SCSI queue depth on legacy AVIO and NPIV devices
During high I/O load, tools like glance, when run inside a vPar or VM shows a very large value
against the Qlen field. Qlen is an indication of number of I/Os that in queue waiting to be processed
by the device. One way to reduce this is to tune the SCSI Queue Depth on the guest devices. This
value is the maximum number of concurrent I/O requests that could be outstanding for a device
and it must be based on the capability of the actual physical device to which the guest device is
mapped to on the VSP. The SCSI queue depth can be set or viewed on a vPar or VM device using
the scsimgr command, just the way it is set or viewed on a physical server.
For more information about tuning the SCSI queue depth for AVIO devices, see HP Integrity VM
Accelerated Virtual I/O (AVIO) Overview white paper at http://www.hp.com/go/hpux-hpvm-docs.
A.3 Networking
A.3.1 AVIO networking
Do not kill hpvmnetd
Do not use the kill command to remove the hpvmnetd process. The following error message
indicates that the hpvmnet daemon has been killed:
hpvmnetd: Switch 0000564d4c414e31 already exists
If the hpvmnetd process is removed, vswitches do not work properly.
AVIO LAN devices not claimed by guest with DOWN vswitch at boot time
In addition to running ioscan, it is necessary to re-run network startup scripts so that IP addresses
can be configured on network interface cards (NICs). For example:
/sbin/rc2.d/S340net start
/sbin/rc2.d/S340net-ipv6 start
A.3 Networking 269