FCoE boot from UCS B200

[Kobus Raath]

Hi

I am running a test environment consisting of a UCS --(fc)--> MDS --(fc)--> Storage.

In an attempt to boot from SAN I see some issues on the MDS.  Although I am able to see the flogi from both the server and the storage, and both are zoned correctly the fcns database does not seem to be correct...  I see the storage at “scsi-fcp:target” but I do NOT see the sever as “scsi-fc:init” – this field is blank.

MDS-1#

MDS-1# sho zones ac

zoneset name zs_vsan100 vsan 100

  zone name zone_vsan100 vsan 100

  * fcid 0xec0003 [pwwn 50:00:1f:e1:50:09:0c:89]

    pwwn 20:00:00:25:b5:00:00:01

  * fcid 0xec0004 [pwwn 20:00:00:25:b5:00:00:02]

    pwwn 20:00:00:25:b5:00:00:03

    pwwn 20:00:00:25:b5:00:00:04

MDS-1#

MDS-1# sho flog dat

---------------------------------------------------------------------------

INTERFACE  VSAN    FCID            PORT NAME               NODE NAME      

---------------------------------------------------------------------------

fc1/2      100   0xec0001  20:42:00:05:73:d8:80:c0  20:64:00:05:73:d8:80:c1    

fc1/2      100   0xec0004  20:00:00:25:b5:00:00:02  20:00:00:25:b5:00:00:03    

fc1/10     100   0xec0003  50:00:1f:e1:50:09:0c:89  50:00:1f:e1:50:09:0c:80    

MDS-1#

MDS-1# sho fcns dat

VSAN 100:

--------------------------------------------------------------------------

FCID        TYPE  PWWN                    (VENDOR)        FC4-TYPE:FEATURE

--------------------------------------------------------------------------

0xec0001    N     20:42:00:05:73:d8:80:c0 (Cisco)         npv

0xec0003    N     50:00:1f:e1:50:09:0c:89 (HP)            scsi-fcp:target 244

0xec0004    N     20:00:00:25:b5:00:00:02                

MDS-1#

If I boot the server from CD (esx installer) I do see the sever as “scsi-fc:init” but only after esx is loaded.  Although it appears to be correct at this point I am still not able to access the LUN from the esx installer.

MDS-1#

MDS-1# sho fcns dat

VSAN 100:

--------------------------------------------------------------------------

FCID        TYPE  PWWN                    (VENDOR)        FC4-TYPE:FEATURE

--------------------------------------------------------------------------

0xec0001    N     20:42:00:05:73:d8:80:c0 (Cisco)         npv

0xec0003    N     50:00:1f:e1:50:09:0c:89 (HP)            scsi-fcp:target 244

0xec0004    N     20:00:00:25:b5:00:00:02                 scsi-fcp:init fc-gs

From this test it looks like a boot driver in the UCS that does not initiate correctly as well as a further issue on the storage array.  Where would I look to resolve this issue?

Regards

[Walter Dey]

Is UCS running in FC End Host Mode (NPV)

Is MDS configured for NPIV

Do you see the vHBA doing a flogi (Initiator pwwn)

Is the Initiator pwwn used for LUN masking / mapping

Are you using different VSAN's for fabric A and B ? and if not why ?

[Kobus Raath]

Hi

To answer the questions please see comments below.

Is UCS running in FC End Host Mode (NPV)

Yes, the UCS is running in End Host mode.

Is MDS configured for NPIV

Yes, the MDS is configured for NPIV

Do you see the vHBA doing a flogi (Initiator pwwn)

Yes, I can see the vHBA do a flogi and is seen the flogi database.  When I look at the fcns database though, it shows but do not have the “scsi-fcp:init fc-gs” suffix.  Only if I boot from HDD and ESX completes loading I see the initiator wwn as “scsi-fcp:init fc-gs”.  See output further down.

Is the Initiator pwwn used for LUN masking / mapping

 I have had a long chat with our storage chaps and the confirmed that the correct wwns are included for this profile on the array.

Are you using different VSAN's for fabric A and B ? and if not why ?

Yes I am using different vsan’s.  I only showed one – both do the same.

MDS-1# sho fcns  dat   (with the server boot from SAN)

VSAN 100:

--------------------------------------------------------------------------

FCID        TYPE  PWWN                    (VENDOR)        FC4-TYPE:FEATURE

--------------------------------------------------------------------------

0xec0000    N     20:43:00:05:73:d8:80:c0 (Cisco)         npv

0xec0001    N     20:42:00:05:73:d8:80:c0 (Cisco)         npv

0xec0003    N     50:00:1f:e1:50:09:0c:89 (HP)            scsi-fcp:target 244

0xec0004    N     20:00:00:25:b5:00:00:02                

Total number of entries = 4

VSAN 200:

--------------------------------------------------------------------------

FCID        TYPE  PWWN                    (VENDOR)        FC4-TYPE:FEATURE

--------------------------------------------------------------------------

0xea0000    N     20:44:00:05:73:d8:56:40 (Cisco)         npv

0xea0001    N     20:43:00:05:73:d8:56:40 (Cisco)         npv

0xea0002    N     50:00:1f:e1:50:09:0c:8d (HP)            scsi-fcp:target 244

0xea0003    N     20:42:00:05:73:d8:56:40 (Cisco)         npv

0xea0004    N     20:41:00:05:73:d8:56:40 (Cisco)         npv

0xea0006    N     20:00:00:25:b5:00:00:0d                

Total number of entries = 6

MDS-1#

[Walter Dey · Accepted solution]

OK, perfect !

The zoning is done with vhba initator pwwn and storage target ?

If you don't see the lun, it must be the lun masking / mapping; your fcns/flogi is correct, zoning as well.

Don't trust the storage guys :-)

Your concern about fcip init Suffix will be resolved, this is not your Primary Problem.

[Kobus Raath]

Thanks Walter

I’ll check the lun masking.  If this is the issue and gets sorted, will this resolve the fcoe boot problem?

[Walter Dey]

Yes, if you can see the remote lun, and install the OS (ESXi), the next and final step will be the boot process. If after successful OS Installation boot is not working, something is wrong with the boot policy.

[Kobus Raath]

It looks like is a storage problem.  We had an issue that the storage array in our lab (old eol kit) overheated and crashed.  Since the rebuild it is visible on the network but does not seem to present luns to any (not just mine) connected servers.  I’ll have to find some other array to test with.

Thanks for the feedback regarding my query though. 

K