replace 6296UP with a 6248UP

[ThijsVanSeveren]

Hi All,

Situation : I was upgrading a 2x6296UP system with all C-series servers but during that upgrade the (inactive) FI-B died (yes these are old boxes!)  => console output showed lots of 'panic' and 'dump' traces

Problem : i don't have a spare 6296UP, but i do have multiple 6248UP's available

Question : can i replace the dead 6296 with a 6248 ?  i only need about 26 ports, so this should fit on a 6248.  Once the 'new' 6248 is up and running i also want to replace the primary 6296 with a 6248

What i tried (but failed) :
1) i installed the same firmware bundle on the 'new' 6248 and upgraded the infrastructure firmware using this bundle
2) i cleared the config on the 6248, connected the heartbeat ports from the 6296 > 6248, booted the 6248
3) the 6248 detected the 6296 and asked me if it should join the cluster > i answered yes, but then i got this :

Installer has determined that the peer Fabric Interconnect is running a different firmware version than the local Fabric. Cannot join cluster.
Local Fabric Interconnect
UCSM version : 4.2(3h)
UCSM SP version : 4.2(3)SP0
Kernel version : 5.0(3)N2(4.23f)
System version : 5.0(3)N2(4.23f)
System SP version:
local_model_no : UCS-FI-6248UP

Peer Fabric Interconnect
UCSM version : 4.2(3h)
UCSM SP version : 4.2(3)SP0
Kernel version :
System version :
System SP version:
peer_model_no : UCS-FI-6296UP

Peer Fabric Interconnect Model is UCS-FI-6296UP and different from Local Fabric Interconnect Model UCS-FI-6248UP . Cannot trigger the firmware Synchronization

Is it even possible to do this ?  
When i run a 'show version' on both systems i do see a difference :
6248 : kickstart: version 5.0(3)N2(4.23f)   /   system: version 5.0(3)N2(4.23f)
6296 :  kickstart: version 5.0(3)N2(3.22c)   /   system: version 5.0(3)N2(3.22c)

all help is welcome !

[balaji.bandi]

I think that is possible since its only port-capacity, and you mentioned you need only required ports

why not - any way old one dead, so you do not have option to restore the services with available tin in place.

if you want to join that should be same model due to port capacity.

[ThijsVanSeveren]

Hi BB, thanks for your reply !
So "if you want to join that should be same model due to port capacity" means that i wont be able to add the 9248 to the existing cluster, right ?
Would it be possible to save the config of the current 6296 and restore that on a 6248 and then move my servers from the 6296>6248 one by one ?   (i think i can try this with 1 server)

[Riaan van Niekerk]

Yes, a backup and restore of a "logical configuration" XML backup file, with "Preserve Identities" enabled. From

https://www.cisco.com/en/US/docs/unified_computing/ucs/sw/gui/config/guide/141/UCSM_GUI_Configuration_Guide_141_chapter43.html#concept_B5391D74F471417299EECB50165FCD34 (old document but still applies)

"Logical configuration—An XML file that includes all logical configuration settings such as service profiles, VLANs, VSANs, pools, and policies. You can use the file generated from this backup to import these configuration settings to the original fabric interconnect or to a different fabric interconnect. You cannot use this file for a system restore."

We have used this method to split an FI-6248 domain to two separate FI-6248 domains (one for prod, one for non-prod) a couple of years ago.

[ThijsVanSeveren]

Update :
After fixing errors related to certificates and encryption I keep hitting this error :
"Failed to instantiate a new instance of the class under MO:fabricEthEstcCloud, DN:fabric/eth-estc, if user action is already in progress"

The www keeps directing me to storage related issues, but I cant find anything wrong with my config.
I also tried importing other backup files (All config / Logical config - both with/without 'preserve identities') but i hit the same issue.  I even tried importing via CLI but that doesnt make any difference.

There is 1 more active warning on the system :
   ​Affected object : sys/import-config-10.50.233.214
   Description : [FSM:STAGE:FAILED]: Report results of configuration application (FSMSTAGE:sam:dme:MgmtImporterImport:reportResults)
   Cause : report-results-failed
   Code : F16684

But that doesnt help much

Any ideas ?

[dbasilio]

Have you tried to delete the failed import? 

scope system
 delete import-config 10.50.233.214
 commit-buffer

[ThijsVanSeveren]

Hi Dbasilio,
yes i did : for each test i did i deleted the previous import and created a new one.  I tried the different types of backup files just to see if any of these work (both via the GUI and via CLI) => none of the imports work 

I had a look at the config file and tracked down the section that seems to cause the problem (MO:fabricEthEstcCloud, DN:fabric/eth-estc) and i removed that section just to see what would happen ... but then the problem moves to an other section.  If i remove that section it moves further etc etc ...
So it feels like there is a problem with the xml format or version, but I cant find any version reference in the XML

I wonder if this is related to the version diff i mentioned in my original post ?