changing preferred path for rdm

[WestleyTDPR]

I have several luns attached as rdm's and can edit the vm settings to set the paths, but it doesn't swap path when i tell it to change,

also if i vmotion the vm to another host the rdm's change back to most recently used path from the fixed setting i had set!

Can anyone enlighten me to whichever glaringly obvious point i'm overlooking

[WestleyTDPR]

....bump.....

[kastlr]

Hi,

not quite sure if I got you right.

Usually, ESX does choose the proper failover policy for the attached array.

These settings would be

- MRU for an active/passive array

- Fixed for an active/active array

It doesn't care if you're using the LUN with or without a VMFS on it.

And the failover policy is set on each ESX host and not part of the VM configuration.

Therefor you should verify if the failover policy for a shared LUN is consistent on all ESX servers.

Each LUN with multiple pathes could only assigned to a VM by selecting the device identifier for that LUN.

I.e. the following output is taken from an A/P array and the LUN is connected via 4 pathes to the ESX host

Disk vmhba0:0:19[/b] /dev/sds (131072MB) has 4 paths and policy of Most Recently Used

- FC 1:6.0 210000e08b1ab44a<->500601611020df1d vmhba0:0:19 Standby preferred

- FC 1:6.0 210000e08b1ab44a<->500601691020df1d vmhba0:1:19 On active

- FC 1:8.0 210000e08b196a7b<->500601601020df1d vmhba1:0:19 Standby

- FC 1:8.0 210000e08b196a7b<->500601681020df1d vmhba1:1:19 On

To use this LUN as RDM in a VM, you must add vmhba0:0:19[/b].

None other "possible" path will be accepted, when I remember right, ESX is that "smart" and will replace it with vmhba0:0:19[/b].

Hope this helps a bit.

[WestleyTDPR]

I'll try to clarify my situation

For load balancing this cluster of four ESX boxes has the preferred path staggered for all shared storage with the first host set to Fixed and the remaining three set to MRU, this is to prevent thrasing from occuring when a path returns. (maybe the fixed settings should be spread in case of host failure but lets ignore that for now)

The RDM attached to the VM have their preference and path set within the vm settings on a per hard disk basis, and this is where i have set them.

As they are only accessed by one machine i'm not worried about thrashing but the SAN man complains that he gets warnings for each lun not on the SAN's preferred path.

I have set them to fixed and set the preferred path, its just that the current path hasn't switched to be the preferred path, there are no messages or warnings from esx

[kjb007]

Can you post your esxcfg-mpath -l before and after you run your command for the LUN you are concerned with?

Also, do you see any errors in your vmkernel and vmkwarning logs?

-KjB

[Natsidan]

Have you done a re scan since you changed the fixed path. Is your san Active/Active or A/P when you do a rescan you can run some cli

watch tail /var/log/vmkernel log file to see what is going on. Path thrashing is more likely to occur with a mixture of MFU and fixed path policies

we are all on a journey