Inconsistent SATP on hosts with same hardware between clusters

[Jonathan77]

Hi,

    we have multiple cluster that have several hosts.

Host Version : ESXi 5.0 U3

Storage : SVC with DS8800 and XIV

The issue we have is :

Let say we have Cluster A, Cluster B, etc..

All hosts on Cluster A uses : VMW_SATP_SVC, they are set to Round Robin, and all of the 4 Paths are Active IO.

All hosts on Cluster B uses : VMW_SATP_ALUA and we set it to Round Robin, and it use 2 paths with Active IO.

We have some other clusters that uses one or the other SATP...

The question is what can cause the hosts to use ALUA instead of SVC...
We want to use SVC and have all the path Active IO...

Based on what I found on SVC best practices is that ESXi 5.0 / 5.1 are supposed to use the SVC by default and 5.5 the ALUA.

We have 5.0 wall to wall and that mixed up SATP is very strange.

Any input on this would be appreciated.

Jon.

[RyanH84]

Hi, is the VMW_SATP_SVC SATP available on the Host B hosts?

esxcli storage nmp satp list

If not, then that could be the reason why. Failing that, do you have any SATP rules in place that might be setting the VMW_SATP_ALUA SATP for your SVC device?

esxcli storage nmp satp rule list

Could have some key software been installed in the first cluster for SVC but not the second to allow this? Were they built at the same time with the same ESXi image. Have their updates differed at all?

You could do a comparison of installed software packages on the hosts to see if something is missing?

esxcli software vib list

Just some random thoughts I had when reading.

Happy to help further if my ramblings have been any use!

Ryan

[Jonathan77]

looks like that changed after we updated the emulex driver.

Since we are dealing with another issue on storage and will be updating our host to 5.5 soon, if it is not causing any issues, then we will let it run like that for now, instead of making any changes.

Thanks for your reply