Only one HBA used in Active/Active with fixed path selection

[Remko_N]

I have 2 ESX hosts with 2 FC HBA's each. They connect to an Active/Active HP EVA 4000 SAN.

I've configured fixed paths to balance the load between the HBA's and EVA Controllers.

For example:

HBA 0 connects to LUN 11 on Controller 0 (vmhba0:0:11)

HBA 1 connects to LUN 12 on Controller 1 (vmhba1:1:12)

But when I look in VC under Storage (SCSI, SAN, and NFS) in the Storage screen, every Datastore is listed as "vmhba0:0:x:x", so it seems that only HBA0 is used, connecting to Controller 0. Clicking on the properties of a datastore, for example LUN12, and choose "Manage Paths", the active path however is "vmhba1:1:12". All other paths are listed as Standby.

When I clone a VMDK from LUN 11 to LUN 12, looking with ESXTOP I never see any usage of vmhba1; only vmhba0.

Here's the output of esxcfg-mpath -l

\[root@esx-z2 root]# esxcfg-mpath -l

RAID Controller (SCSI-3) vmhba0:0:0 (0MB) has 4 paths and policy of Fixed

FC 5:0.0 10000000c96461d7<->50001fe1500d488c vmhba0:0:0 On active preferred

FC 5:0.0 10000000c96461d7<->50001fe1500d4888 vmhba0:1:0 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d488d vmhba1:0:0 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d4889 vmhba1:1:0 On

Disk vmhba0:0:10 /dev/sda (204800MB) has 4 paths and policy of Fixed

FC 5:0.0 10000000c96461d7<->50001fe1500d488c vmhba0:0:10 On

FC 5:0.0 10000000c96461d7<->50001fe1500d4888 vmhba0:1:10 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d488d vmhba1:0:10 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d4889 vmhba1:1:10 On active preferred

Disk vmhba0:0:11 /dev/sdb (512000MB) has 4 paths and policy of Fixed

FC 5:0.0 10000000c96461d7<->50001fe1500d488c vmhba0:0:11 On active preferred

FC 5:0.0 10000000c96461d7<->50001fe1500d4888 vmhba0:1:11 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d488d vmhba1:0:11 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d4889 vmhba1:1:11 On

Disk vmhba0:0:12 /dev/sdc (307200MB) has 4 paths and policy of Fixed

FC 5:0.0 10000000c96461d7<->50001fe1500d488c vmhba0:0:12 On

FC 5:0.0 10000000c96461d7<->50001fe1500d4888 vmhba0:1:12 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d488d vmhba1:0:12 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d4889 vmhba1:1:12 On active preferred

Disk vmhba2:0:0 /dev/cciss/c0d0 (69973MB) has 1 paths and policy of Fixed

Local 70:0.0 vmhba2:0:0 On active preferred

Seems to me everything is configured OK. A reboot doesn't help either.

What am I doing wrong here?

[Erik_Zandboer]

Hi, As far as my knowledge goes, ESX only supports active/passive SAN connections. This might be the problem as to why you see only a single active path.

[Texiwill]

Hello,

ESX supports Active/Active SANs yet it does not load balance or agregate the paths. It supports basically failover so balancing an active/active SAN or more than one SAN is left up to the administrator.

Is this a boot from SAN configuration?

RAID Controller (SCSI-3) vmhba0:0:0 (0MB) has 4 paths and policy of Fixed

FC 5:0.0 10000000c96461d7<->50001fe1500d488c vmhba0:0:0 On active preferred

Disk vmhba0:0:10 /dev/sda (204800MB) has 4 paths and policy of Fixed

FC 67:0.0 10000000c94e5ecd<->50001fe1500d4889 vmhba1:1:10 On active preferred

Disk vmhba0:0:11 /dev/sdb (512000MB) has 4 paths and policy of Fixed

FC 5:0.0 10000000c96461d7<->50001fe1500d488c vmhba0:0:11 On active preferred

Disk vmhba0:0:12 /dev/sdc (307200MB) has 4 paths and policy of Fixed

FC 67:0.0 10000000c94e5ecd<->50001fe1500d4889 vmhba1:1:12 On active preferred

This actually shows that things are load balanced properly. vmhba0 for half your paths and vmhba1 for the other half. VC is basically out of date you can 'refresh' the screen in the VIC or from the command line run:

service vmware-mgmt restart

service vmware-vpxa restart

And reconnect your VIC. That should fix the problem.

BTW, the original EVA 4K is an active/passive SAN. That is unless HP came out with new Firmware to make it active/active.... You may wish to check on this, as otherwise you are slamming your SAN with constant Storage processor switching.

Best regards,

Edward

[VirtualKenneth]

BTW, the original EVA 4K is an active/passive SAN.

That is unless HP came out with new Firmware to make

it active/active.... You may wish to check on this,

Yep, EVA4K's with a new VCS (or is it XCS code are passive/passive.

(even the EVA3K / 5K can be passive/passive with new code)

[BUGCHK · Accepted solution]

>> BTW, the original EVA 4K is an active/passive SAN.

>Yep, EVA4K's with a new VCS (or is it XCS code are passive/passive.

Wrong. Wrong. Wrong. The EVA4000 has always been active/active and the firmare code is nick-named XCS[/b] (eXcellerated Controller Software). Beleive me. I was at the EVA engineering labs in the Colorado Springs facilities a few weeks before the EVA4000, 6000 and 8000 were officially announced and visited an introduction session.

(even the EVA3K / 5K can be passive/passive with new code)

The EVA3000 and EVA5000 have been active/passive from day one (I started with VCS V1.002 - the EVA3000 was introduced around VCS V2.003 or so; sorry, too lazy to check right now). Active/active was introduced with the VCS Version 4 codestream, but Version 3 is supported as well.

Remko N:

the display of the datastore is always associated with the '[u]cannonical name[/u]'. That is why you need to check, e.g. with esxcfg-mpath[/b]

Disk vmhba0:0:12 /dev/sdc (307200MB) has 4 paths and policy of Fixed

FC 5:0.0 10000000c96461d7<->50001fe1500d488c vmhba0:0:12 On

FC 5:0.0 10000000c96461d7<->50001fe1500d4888 vmhba0:1:12 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d488d vmhba1:0:12 On

FC 67:0.0 10000000c94e5ecd<->50001fe1500d4889 vmhba1:1:12 On active preferred

In this example, vmhba0:0:12 is the cannonical path, but vmhba1:1:12 is the path used for I/O.

So the 4 virtual disks are balanced across both controllers, although you only use one controller's Port_2 and the other controller's Port_1, but I don't think you are \_that_ bandwidth limited.

[Texiwill]

Hello,

You know I was confused, comparing it to a 3K in my mind....

Best regards,

Edward