cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1720
Views
0
Helpful
7
Replies

UCS blade refresh M3 to M5 EVC

joeharb
Level 5
Level 5

We are in the process of refreshing M3 Blades with M5 at 2 different locations.  The first location had 2 B200 M3 blades in a ESXi Cluster, we were able to replace the first blade with a M5 and attach the original M3 service profile and bring the bade into service within Vcenter.  All VM's were then migrated from the M3 to the M5 using Sandy Bridge EVC.  Then the other M3 blade was replaced, vm's separated between the 2 servers without issue.

 

Second site had 2 M3 blades and only 1 of them (off the books) is being replaced with a M5 blade.  Did the same process, moved all vm's M3, replaced the M3 with M5 and assigned service profile.  Vcenter will not allow for the migration of vm's to the M5 due to incompatibility.  Vcenter will not allow for the selection of EVC Sandy bridge as it states it is not compatible.  I am verifying the same firmware is being used, but I am not finding anything that would explain why one site allowed the the blades to coexist but the other will not.  Vcenter version is the same on both sites.

 

Please advise,

 

Joe 

7 Replies 7

Kirk J
Cisco Employee
Cisco Employee

Greetings.

Is there any difference in the Procs in use on the M3 or M5 blades in your 2nd site compared to your first site?

Are there any additional BIOS policies applied on the blades in site 2?

Maybe try decoming and re-acking the blades in site 2 ?

Does it say if it's complaining about your M3s or M5s for the evc issue?

 

Kirk...

We reinserted a blade in both location and the proc's are the same:

 

One thing that I did notice is that at site 1 one of the servers bios policy has an older version applied to the BIOS...I am not completely sure which one was changed out first, but I would expect this one was replaced first.  So in site 1 both M5 are running 2 different firmware packages.  

Blade one is package 3.2(3a)B the other is 4.0(1a)B.

In Site 2 both the M3 and M5 are running 4.0(1a)B.

 

Should I attempt to change the policy on the M5 in site 2 the lower version to see if that allows for vmotion?

 

Thanks,

 

Joe

 

 

 

 

 

Wonder if the Pre vs Post fix for Intel spectre/meltdown related type setting are coming into play?

Will be curious if the firmware downgrade resolves the evc message...

 

Kirk...

Downgrade of firmware did not resolve the issue.

Maybe try running vim-cmd hostsvc/hosthardware > /tmp/hardwareinfo.txt for each of the hosts in question, and then a compare app against the CPU / CPU features sections of the output to see if something jumps out as significantly different between site 1 and site 2

Kirk...

I am just looking at this again, maybe a fresh look helped.  I checked the original site and it is configured for Sandy Bridge EVC.  It was set this way prior for a M2 to M3 migration.  So I assume that is what allowed for the introduction of the M5 to the cluster with M3's and vmotion to work without issue.

Since this cluster has only M5 cards in it should I disable EVC completely or change to a more current Mode?


I am assuming at this point that I can got ahead and disconnect/remove the M5 from the cluster that I am having issues with and change the EVC to Sandy bridge then add it back, I am assuming this will allow for Vmotion between the servers...

 

Is my logic correct?

 

Thanks,

 

Joe

Any luck on this ? Have you managed to get your VMs vmotioned on the M5 blade ? have you tried rebooting your VMs ?

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: