DS4800 Reboot / Rescan Losing Preferred Path

[tadsmith]

I have a number of HS21 blades, each with 2 HBAs, connected to a DS4800 over seperate fabrics for each controller (A and B). ESX is set to MRU multipathing. Whenever I reboot or rescan the HBAs, a number of the connected LUNs go to a non-preffered path. I have to manually go in and revert the paths back using IBM Storage Manager. This can be quite cumbersome when updating/patching a large number of hosts. Does anyone have any ideas/suggestions for dealing with this?

[Khue]

Sounds like a zoning problem possibly. How many HBA ports are there on each node of the 4800? Are there 2 on each ESM?

[tadsmith]

I don't believe it's a zoning issue. Each server sees the following 4 paths:

HBA1 -> DS4800 A1

HBA1 -> DS4800 A3

HBA2 -> DS4800 B1,

HBA2 -> DS4800 B3

All of the paths work, it's just that when the HBAs rescan, they move some of the LUNs off the preferred path (from A to B or vice versa). Since it's set to MRU, I have to manually go in and "fail-back" the LUNs each time. It can be a hassle having to reboot the server, fail-back the luns, patch, reboot, fail-back, etc, especially for multiple hosts.

[Khue]

In storage manager (I am on storage manager 10), click configured hosts. How many wwpns do you see there relating to the server you are having issues with?

[tadsmith]

Like I said, all paths work, they just tend to flip back and forth on reboot/rescan. They still run, just not on the preferred path.

[Khue]

Ok, good luck figuring it out. Sorry I couldn't be of more help. I do know that I have had similar problems on a DS4700, DS3300, and a DS3400 where the pathing was correct, however not enough paths were presented. Each host port needs to be able to see each subsystem port no matter what. There are a total of 8 paths between each of my servers and subsystems.

[thecakeisalie]

Run "esxcfg-mpath -lv" on all of your hosts, and if you can't find the issue then save the output from all of them to a file and upload it here.

[homerjnick]

I thought this was a known issue with VMWare since you cant install any IBM RDAC drivers into ESX and there is nothing you can do about it?

We had this issue all the time on our DS units but we just moved to a Netapp SAN in our Production environment and they have drivers for multipathing so my issue has gone but in our test environment where we use 3 DS4300 units we get this all the time.

[tadsmith]

Here is a sample.

esxcfg-mpath -lv | grep "Fixed\|preferred"

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:16 Standby preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:14 Standby preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:2 Standby preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:12 Standby preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:0 Standby preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:3 On preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:13 On active preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:15 On active preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:4 On preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:5 Standby preferred

FC 8:1.0 210000e08b9da78e:200000e08b9da78e<->201400a0b826a1b8:200400a0b826a1b8 vmhba2:0:1 On active preferred

It looks like there is a descrepency between the preferred path on the SAN side and the preferred path on the ESX side. Is there a way to clear the preferred paths from ESX and rebuild them? I just tried a server and the only way I could find to do it was quite cumbersome. I had to change the policy to fixed, change the active/preffered path, and then change the policy back to MRU for each LUN.

[homerjnick]

"It looks like there is a

descrepency between the preferred path on the SAN side and the

preferred path on the ESX side. Is there a way to clear the preferred

paths from ESX and rebuild them? I just tried a server and the only way

I could find to do it was quite cumbersome. I had to change the policy

to fixed, change the active/preffered path, and then change the policy

back to MRU for each LUN."

Thats what we have to do.

[ldean]

We had the same issue here with our DS4800 and ESX 4.0. My solution was to switch to Hyper-V which fixed the issue quite nicely. It seems the VMWare boys couldn't help me at the time. Hopefully there is a fix to this otherwise, Hyper-V, here we come. I have to admit SCVMM R2 is coming along nicely.

Chris