The setup of NPIV is a little unusual, in that the VMware guest needs to be initially powered off. You then configure fabric zoning and array LUN masking using the NPIV WWNs whilst not currently logged in to the fabric. You then power on the guest, and ESX will ‘bring up‘ the NPIV virtual N_Ports/WWNs. These virtual N_Ports will then FLOGI/FDISC and PLOGI, and FC4 will then attempt to inquire about LUNs behind the storage target (PRLI). At this point, if a LUN cannot be seen, the WWNs will log out (it is worth remembering that it is a requirement of the SCSI protocol that every SCSI target should respond to an inquiry against LUN zero).
So, if you are troubleshooting NPIV WWNs that are absent from the fabric nameserver, power-cycle the guest and keep issuing either a 'show flogi database' or an ‘nsshow‘ command - if you see the WWNs for a few seconds only, then LUN provisioning is quite likely your problem. Basically, provision the zoning and LUN masking ahead of time (physical and virtual), and then the WWNs should not log out.
A common error that I have seen is to provision the RDM device to the ESX server, this then ‘toggles on‘ the ability to generate virtual N_Port WWNs, however, administrators can then subsequently LUN mask LUNs that have not been provisioned as RDM devices.
felipon1
96 Posts
0
May 22nd, 2012 04:00
Hi,
I think that your missing some steps. Please, take a look at the following doc about it:
http://blogs.vmware.com/vsphere/2011/11/npiv-n-port-id-virtualization.html
Rgds
Richard_Butler
84 Posts
0
June 22nd, 2012 01:00
Hello,
The setup of NPIV is a little unusual, in that the VMware guest needs to be initially powered off. You then configure fabric zoning and array LUN masking using the NPIV WWNs whilst not currently logged in to the fabric. You then power on the guest, and ESX will ‘bring up‘ the NPIV virtual N_Ports/WWNs. These virtual N_Ports will then FLOGI/FDISC and PLOGI, and FC4 will then attempt to inquire about LUNs behind the storage target (PRLI). At this point, if a LUN cannot be seen, the WWNs will log out (it is worth remembering that it is a requirement of the SCSI protocol that every SCSI target should respond to an inquiry against LUN zero).
So, if you are troubleshooting NPIV WWNs that are absent from the fabric nameserver, power-cycle the guest and keep issuing either a 'show flogi database' or an ‘nsshow‘ command - if you see the WWNs for a few seconds only, then LUN provisioning is quite likely your problem. Basically, provision the zoning and LUN masking ahead of time (physical and virtual), and then the WWNs should not log out.
A common error that I have seen is to provision the RDM device to the ESX server, this then ‘toggles on‘ the ability to generate virtual N_Port WWNs, however, administrators can then subsequently LUN mask LUNs that have not been provisioned as RDM devices.
Best regards, Richard.