EVA 4400 and 11.31 Multipathing

[No idea what I'm doing]

Hey, guys. I've got two rx3600 Itanium servers running HP-UX 11.31 and they are clustered together in an active/passive configuration via ServiceGuard. The servers have two dual port combo cards in them for a total of 4 HBA ports per server and they are both connected to an EVA 4400.

I'm noticing that I only see activity across two of the four HBA ports:

root@rachp33:/var/adm/sa# sar -H -f sa01

HP-UX rachp33 B.11.31 U ia64 03/01/09

00:00:00 ctlr util t-put IO/s r/s w/s read write avque avwait avserv
%age MB/s num num num MB/s MB/s num msec msec
00:05:00 fcd1 32 7.71 245 188 57 5.36 2.35 1 0 2
sasd0 79 3.74 451 2 449 0.01 3.73 1 0 7
fcd3 32 7.72 245 188 57 5.34 2.38 1 0 2
00:10:00 fcd1 20 5.53 120 84 36 3.62 1.90 1 0 3
sasd0 3 0.13 10 0 10 0.00 0.13 1 0 12
fcd3 21 5.61 120 84 36 3.70 1.90 1 0 3


I don't see anything for fcd0 and fcd2. I've reset the stats for all four fc devices via fcmsutil. I look at the fcmsutil stats and it's clear that most of the traffic is going over only two of the four ports.

As I understand it, the EVA4400 is an active/active array. I have a round-robin policy set on all of my LUNs presented from the EVA and have done nothing to change that via scsimgr. It's zoned such that each lun has four paths.

Do I need to turn on the preferred path policy and then set the preferred path to the primary controller for that LUN in order to get traffic going over all four HBA ports? I worked with HP a while back and they wanted me to do this. We have since gotten a newer, more robust EVA 4400 and I haven't yet done anything to the LUNS that were presented from the new array.

Thank you,

[Robert-Jan Goossens]

Hi,

How did you configure you lvm structure?

# vgdisplay -v /dev/vgxx

Are all alternate visible?

# ioscan -fnCdisk

Robert-Jan

[Robert-Jan Goossens]

add on;

http://docs.hp.com/en/5992-4589/ch02s03.html#cbdbgbgh

[No idea what I'm doing]

--- Volume groups ---
VG Name /dev/vg44
VG Write Access read/write
VG Status available, exclusive
Max LV 255
Cur LV 5
Open LV 5
Max PV 255
Cur PV 97
Act PV 97
Max PE per PV 1151
VGDA 194
PE Size (Mbytes) 32
Total PE 111647
Alloc PE 108194
Free PE 3453
Total PVG 1
Total Spare PVs 0
Total Spare PVs in use 0
VG Version 1.0
VG Max Size 9392160m
VG Max Extents 293505

I have 97 LUNs in the VG so I won't paste the entire output. All physical volumes are in one physical volume and I use a PVG strict allocation policy on the LVs.

Here's a sample of ioscan -fnNkC disk

disk 62 64000/0xfa00/0x9 esdisk CLAIMED DEVICE HP HSV300
/dev/disk/disk62 /dev/rdisk/disk62
disk 63 64000/0xfa00/0xa esdisk CLAIMED DEVICE HP HSV300
/dev/disk/disk63 /dev/rdisk/disk63
disk 64 64000/0xfa00/0xb esdisk CLAIMED DEVICE HP HSV300
/dev/disk/disk64 /dev/rdisk/disk64
disk 65 64000/0xfa00/0xc esdisk CLAIMED DEVICE HP HSV300
/dev/disk/disk65 /dev/rdisk/disk65
disk 66 64000/0xfa00/0xd esdisk CLAIMED DEVICE HP HSV300
/dev/disk/disk66 /dev/rdisk/disk66
disk 67 64000/0xfa00/0xe esdisk CLAIMED DEVICE HP HSV300
/dev/disk/disk67 /dev/rdisk/disk67
disk 68 64000/0xfa00/0xf esdisk CLAIMED DEVICE HP HSV300


Here is the output of scsimgr lun_map -D /dev/disk/disk900

root@rachp33:/root# scsimgr lun_map -D /dev/rdisk/disk960 | more

LUN PATH INFORMATION FOR LUN : /dev/rdisk/disk960

Total number of LUN paths = 4
World Wide Identifier(WWID) = 0x600508b4000b2e760000f00001f80000

LUN path : lunpath557
Class = lunpath
Instance = 557
Hardware path = 0/7/1/0/4/1.0x50014380025d1bc9.0x4061000000000000
SCSI transport protocol = fibre_channel
State = ACTIVE
Last Open or Close state = ACTIVE

LUN path : lunpath609
Class = lunpath
Instance = 609
Hardware path = 0/7/1/0/4/1.0x50014380025d1bcd.0x4061000000000000
SCSI transport protocol = fibre_channel
State = STANDBY
Last Open or Close state = STANDBY

LUN path : lunpath748
Class = lunpath
Instance = 748
Hardware path = 0/3/1/0/4/1.0x50014380025d1bc8.0x4061000000000000
SCSI transport protocol = fibre_channel
State = ACTIVE
Last Open or Close state = ACTIVE

LUN path : lunpath800
Class = lunpath
Instance = 800
Hardware path = 0/3/1/0/4/1.0x50014380025d1bcc.0x4061000000000000
SCSI transport protocol = fibre_channel
State = STANDBY
Last Open or Close state = STANDBY

[Duncan Edmonstone · Accepted solution]

Well it looks to me like you only have the vdisks presented or zoned to 2 of the 4 FC cards (0/7/1/0/4/1 and 0/3/1/0/4/1).

So you need to check the rest of you're infrastructure:

i) Determin the HW path and device file for all 4 FC ports:

ioscan -funC fc

ii) Check the ports are logged in to your FC SAN fabric:

fcmsutil /dev/fcd0

(using the device files identified in step i)

iii) Assuming they are logged in, check your swithc to ensure the FC WWNs you see from the outputs of fcmsutil are zoned in to the zone with your EVA ports.

iv) Ensure the vdisks are presnted to all 4 FC WWNs from within Command View EVA. You'll need to see all 4 FC ports defined for the host's config.

HTH

Duncan

[No idea what I'm doing]

This was indeed a zoning thing. Thank you for your help, Duncan.