Field Notice: FN74391 - When Adding Large Numbers of VLANs to a vNIC, Redundant Fabric Interconnects May Both Restart - Software Upgrade Recommended

Cisco UCS Fabric Interconnect models UCS-FI-6454, UCS-FI-64108, and UCS-FI-6536 may encounter unexpected reboots on both devices after upgrading to affected releases with allowed vLANs on vEthernets that surpass 900 characters.

The issue occurs when there is a change in the vEthernet status or a configuration change on the vEthernet, including the following:

• Server reboot
• vNIC link flap from Intersight/UCS Manager/operating system
• vLAN addition
• vLAN removal (but still having more than 900 characters)

This can cause a full data and management plane outage for the impacted environment.

This issue arises from a Fatal Module Error that is triggered by iftmc hap reset.

The condition occurs when the allowed vLAN list on an individual vEthernet has more than 900 characters in the string. This includes dashes ( - ), commas ( , ), and any numbers (for example, 1-3,5,7,9-11)

This issue impacts vEthernet only. It will not impact uplink or appliance ports carrying 900 characters worth of vLANs.

Cisco UCS uses Virtual Ethernet (vEthernet) interfaces as the binding interface for virtual NICs (vNICs) on Virtual Interface Cards (VICs).

Both Fabric Interconnects in a Cisco UCS domain could reboot unexpectedly and sequentially, resulting in all connected servers becoming inaccessible during the reboot period. 

When the Fabric Interconnects reboot, they will be inaccessible. All traffic and services going over the affected Fabric Interconnect cluster will be halted. 

This impacts the following Cisco UCS Fabric Interconnect models:

• UCS-FI-6454
• UCS-FI-64108
• UCS-FI-6536

To check if a Fabric Interconnect rebooted because of this issue, SSH to the Fabric Interconnect after it has recovered and run the following:

FI-X# connect nxos
FI-X(nx-os)# show system reset-reason
----- reset reason for module 1 (from Supervisor in slot 1) ---
1) At 351285 usecs after Tue Feb 28 10:00:00 2026
  Reason: Reset Requested due to Fatal Module Error
  Service: iftmc hap reset
  Version: 10.5(1)I60(1e)
FI-X(nx-os)# show logging nvram
2026 Feb 30 10:00:00 FI-X %$ VDC-1 %$ %SYSMGR-SLOT1-2-SERVICE_CRASHED: Service "iftmc" (PID 10193) hasn't caught signal 11 (core will be saved).
2026 Feb 30 10:00:00 FI-X %$ VDC-1 %$ %SYSMGR-SLOT1-2-HAP_FAILURE_SUP_RESET: Service "iftmc" in vdc 1 has had a hap failure

If the Fabric Interconnect sees iftmc hap reset from show system reset-reason on impacted firmware, then the Fabric Interconnect was impacted by this issue. 

Solution

To fix the issue that is described in this field notice, upgrade to a fixed software release as shown in the following table:

Download the latest release of Cisco UCS Infrastructure Software Bundle from the Cisco Software Download Center. 

Until the Fabric Interconnects can be upgraded to a fixed release, reduce the number of vLANs in the vLAN string to fewer than 900 characters. 

To check if a device is susceptible to the defect on impacted firmware, use the following command to output every vEthernet that is present on the Fabric Interconnects:

FI-X# connect nxos
FI-X(nx-os)show run int | grep -A 10 -i Veth
--
interface Vethernet827
 description server 1/1, VNIC eth1
 no lldp transmit
 no lldp receive
 no pinning server sticky
 pinning server pinning-failure link-down
 no cdp enable
 switchport mode trunk
 switchport trunk native vlan 827
switchport trunk allowed vlan 52-58,60,63-75,77-85,87-88,90,96,105-106,108,110-120,122-137,139-151,153,155-157,160,308,311,317,323,350,362,370,372-373,378,382,391,405,413,419-421,425,427-429,432,438,440-442,452,457,469,479,481,492,500-501,505-506,512,528,532,623,700,703-704,706,714,722,736,741-742,747,757-758,763,765,769,772-774,788,790,793,825-827,841,843,857-858,863,866-868,878,889,899,904,921,927,931,940,942,1700,1707,1740-1741,1748,1771,1784,1787,1801,1817,1835,1837,1847,1850,1852,1854,1857,1865,1872,1885,1887,1896,1899,1904-1905,1908,1911,1917,1923-1924,1927,1929,1939,1941-1942,2613,2617,2626,2644,2649,2654,2658,2675,2682,2685,2703,2710,2716,2724,2739,2742,2755,2759,2762,2764,2784,2798,2800,2805-2806,2811,2818,2822,2824,2900,2910-2911,2913,2919,2945-2946,2961,2965,2975,2977,2979,2996,3018,3027,3030,3045,3056,3065,3073,3080,3089,3094,3111,3217,3224,3227-3228,3233,3243,3255,3261,3274,3280-3281,3292,3294,3306,3349,3351,3360,3362,3368,3382-3383,3391,3394,3396,3399,3401,3405,3409,3419,3421,3423,3437,3441,3447,3450,3453-3455,3502-3503,3505,3513,3519-3521,3529,3535,3537,3541-3542,3557,3563-3564,3571-3572,3574-3575,3577-3579,3591,3604,3609,3611,3616,3632,3635,3637-3638,3643-3644,3646,3649,3656,3675-3676,3688,3691,3703,3706,3711,3714,3716,3730,3736,3752,3755
 no hardware vethernet mac filtering per-vlan
 bind interface port-channel1372 channel 849
 no shutdown

This example in the string has 1244 characters in switchport trunk allowed vlan. Character counting for the defect starts at 52 and includes all characters, including dash ( - ) and commas ( , ) and ends at 3755.