I have same lun (hdisk3) assigned to 2 vios server (dual vio server) and this is mapped to the client lpar. I have sddpcm on the VIO.
On the VIOA, the attributes for fscsi1 and fscsi3 are set to fast_fail and dyntrk is set to yes
Now, if I do a lqueryvg -Atp on the hdisk3, I get a "DISK OPERATION ERROR" and a related fscsi1.
I am trying to understand , if this is due to scsi reserve issue or is it a zoning issue. With sddpcm, the hdisk reserve_policy is set to no_reserve by default.
I still have to find out if any errors are reported on the switch.
# lspath -l hdisk3
Enabled hdisk3 fscsi1
Enabled hdisk3 fscsi1
Enabled hdisk3 fscsi3
Enabled hdisk3 fscsi3
# errpt | pg
IDENTIFIER TIMESTAMP T C RESOURCE_NAME DESCRIPTION
DCB47997 0806085511 T H hdisk3 DISK OPERATION ERROR
DCB47997 0806085511 T H hdisk3 DISK OPERATION ERROR
DCB47997 0806085511 T H hdisk3 DISK OPERATION ERROR
4B436A3D 0806085511 T H fscsi1 LINK ERROR
4B436A3D 0806085511 T H fscsi1 LINK ERROR
Not sure if its the same for IBM storage but for EMC + Hitachi, a bug within AIX shows you no reserve despite the fact that the reserve is still on - so you might want to remove the reserve manually again just to be safe.
Regards
zxmaus
As a workaround, I have disabled the fscsi1 path on VIOA server and it I don't see any errors. However, I noticed all I/O's going thru just 1 path (path0) of the adapter (fscsi3) .
I was expecting the I/o to be balanced between Path0 and Path2 for hdisk3 .
Any thoughts ?
VIOA
pcmpath query device
Total Dual Active and Active/Asymmetric Devices : 2
DEV#: 2 DEVICE NAME: hdisk2 TYPE: 2145 ALGORITHM: Load Balance
SERIAL: 600507680190814C7800000000000358
==========================================================================
Path# Adapter/Path Name State Mode Select Errors
0* fscsi3/path2 CLOSE NORMAL 0 0
1 fscsi3/path3 CLOSE NORMAL 0 0
2 fscsi1/path0 CLOSE NORMAL 0 0
3 fscsi1/path1 CLOSE NORMAL 0 0
DEV#: 3 DEVICE NAME: hdisk3 TYPE: 2145 ALGORITHM: Load Balance
SERIAL: 600507680190814C780000000000035D
==========================================================================
Path# Adapter/Path Name State Mode Select Errors
0 fscsi3/path2 OPEN NORMAL 106590 0
1* fscsi3/path3 OPEN NORMAL 0 0
2 fscsi1/path0 OPEN OFFLINE 0 0
3* fscsi1/path1 OPEN OFFLINE 0 0
VIOB
# r
pcmpath query device
Total Dual Active and Active/Asymmetric Devices : 2
DEV#: 1 DEVICE NAME: hdisk1 TYPE: 2145 ALGORITHM: Load Balance
SERIAL: 600507680190814C7800000000000359
==========================================================================
Path# Adapter/Path Name State Mode Select Errors
0* fscsi3/path3 CLOSE NORMAL 0 0
1 fscsi3/path2 CLOSE NORMAL 0 0
2 fscsi1/path0 CLOSE NORMAL 0 0
3* fscsi1/path1 CLOSE NORMAL 0 0
DEV#: 3 DEVICE NAME: hdisk3 TYPE: 2145 ALGORITHM: Load Balance
SERIAL: 600507680190814C780000000000035D
==========================================================================
Path# Adapter/Path Name State Mode Select Errors
0 fscsi3/path2 OPEN NORMAL 24 0
1* fscsi3/path3 OPEN NORMAL 0 0
2 fscsi1/path0 OPEN NORMAL 0 0
3* fscsi1/path1 OPEN NORMAL 0 0
#
actually you cannot really check if the reserve is on or off - but if you do the chdev on devices where it REALLY is off nothing is changing - so no harm done running it after any kind of storage work on both vios (and if you have clusters, on your lpars too)