AIX, VIO, NPIV and VMAX

Reserve_Lock only applies to disks you want to share between LPARS (such as GPFS or Oracle RAC type environments).

I would highly recommend changing the AIX hdisk queue_depth, the default is 3, and should be at least 32 (IMO).

Note that if you have multiple VIOS within a p-series, then you will need to adjust the reserve_lock and possibly the reserve_policy settings. The following extract is from the "EMC Host Connectivity Guide for IBM AIX rev A34":

The System p system can have multiple VIOS LPARs providing access to the same LUN for the same VIOC for Standard Device configurations (no multipathing) and multipathing devices under the control of PowerPath and MPIO. For Standard Device and PowerPath configurations, the Virtual I/O Server device reservation setting reserve_lock must be changed (via the chdev Virtual I/O Server configuration command) to no. For MPIO configurations and PowerPath v5.5 and later, the Virtual I/O Server device reservation setting reserve_policy must be changed (via the chdev command) to no_reserve.

It sounds like he's using NPIV, which negates that comment.

In the old VIO environment, the lun was presented to the LPAR (VIOC) via vscsi (virtual scsi), so the VIOS (server) needed no_reserve enabled to allow redundant VIOS. Thankfully, with NPIV, the fibre channel adapter itself gets virtualized, so the VIOC/LPAR sees a fibre channel adapter, the VIOS cannot actually access the lun, as the storage array masks it directly to the VOIC using a N_port virtual WWN.

This also means that PowerPath can be installed directly on the VIOC/LPAR, as the initial poster described he was doing.

The connectivity guide probably needs a refresh to cover NPIV options, as well the new VIOS7 storage pool options (similar to the VMWare VMFS approach).

Thanks for the clarification - it's hard to keep track of things sometimes, with all the levels of virtualisation on modern systems...

Hi! Thank you for your help. queue_depth is set to 16. I guess it's the default Powerpath value.

# lsattr -El hdiskpower23

clr_q         yes                              Clear Queue (RS/6000)      True
location                                       Location                   True
lun_id        0x3a000000000000                 LUN ID                     False
lun_reset_spt yes                              FC Forced Open LUN         True
max_coalesce  0x10000                          Maximum coalesce size      True
max_transfer  0x40000                          Maximum transfer size      True
pvid          00c0555509f87c7b0000000000000000 Physical volume identifier False
pvid_takeover yes                              Takeover PVIDs from hdisks True
q_err         no                               Use QERR bit               True
q_type        simple                           Queue TYPE                 False
queue_depth   16                               Queue DEPTH                True
reassign_to   120                              REASSIGN time out value    True
reserve_lock  yes                              Reserve device on open     True
rw_timeout    40                               READ/WRITE time out        True
scsi_id       0x11900                          SCSI ID                    False
start_timeout 180                              START unit time out        True
ww_name       0x50000972082cb1e0               World Wide Name            False

I'll try 32. Anything else I should look for?