Cisco Nexus FEX Troubleshooting: FEX Not Coming Online

A Fabric Extender is a remote line card, not a switch: it has no local control plane, no local forwarding decisions and no independent configuration. Every FEX problem therefore has only three possible locations - the fabric ports, the parent switch's association with the FEX, or the host interfaces themselves. Knowing which show command speaks for which layer is most of the diagnosis. This guide collects the states you will see and what they mean.

First look: is the FEX even online?

switch# show fex
FEX         FEX           FEX          FEX
Number      Description   State        Model                Serial
100         FEX0100       Online       N2K-C2248TP-1GE      JAF11223333
101         FEX0101       Offline      N2K-C2232P-10GE      JAF1331AKBM
---         ---           Discovered   N2K-C2232PP-10GE     SSI15500F4Q

• Online - the FEX is associated, image is in sync and host ports are provisioned.
• Offline/Discovered - the FEX is physically present on the fabric link but has no valid association, usually a missing fex associate, a wrong FEX ID or a fabric port that is not in fex-fabric mode.
• Image syncing - the parent is downloading the FEX firmware; normal for a few minutes after a swap.

Fabric layer: pinning, ports and port-channels

switch# show interface fex-fabric
switch# show fex 100 detail
switch# show port-channel summary | include fex

interface Ethernet1/11-12
  channel-group 101 mode on
interface port-channel101
  switchport mode fex-fabric
  fex associate 101

Use a static port-channel (mode on) for fabric links: the parent switch controls the FEX directly, LACP adds nothing and can delay bring-up. Pinning mode (static versus port-channel) determines how host interfaces map to fabric uplinks - with static pinning, an uplink failure takes down specific host interfaces rather than rebalancing, which is why dual-homing exists. Verify the mapping with show fex 100 detail and show fex pinning if your platform exposes it.

# inspect the FEX itself when a parent-side command is not enough
switch# attach fex 100
fex-100# show version
fex-100# exit

Diagnostics on the FEX are read-only; the parent still owns policy. Use show diagnostic result fex 100 for POST results, show environment fex 100 for temperature, power and fan sensors, and show fex 100 transceiver when a fabric link is up but flapping.

Host interfaces: the failures users report

switch# show interface ethernet100/1/1 counters errors
switch# show interface ethernet100/1/1 status
switch# show interface trunk | include 100/1

• Server not connecting at all - wrong FEX port mapping after a cabling change; FEX host ports are numbered chassis/1/port and do not follow the parent's port numbering.
• Errors and drops - tail drops on egress from a FEX uplink occur when several servers burst at once; the FEX buffer is smaller than the parent's. Move bursty servers (storage, backup, video) to parent switch ports, or add fabric uplinks so the FEX has more buffer credit.
• VLAN not passing - the VLAN must exist and be active on the parent switch; a FEX cannot define one. Check show vlan id 123 and the allowed VLAN list on the host interface trunk.
• No Layer 3 on FEX host ports - by design. All routing happens on the parent.

Dual-homed FEX with vPC

A FEX attached to two parent switches must be dual-homed within a single vPC domain, with the fabric interfaces as vPC member ports on both parents. Mismatched VLANs, STP settings or port-channel parameters between the parents break vPC consistency and the FEX drops offline with a consistency failure in the logs. Check show vpc consistency-parameters global and show system internal fex info when the parent's high-level view is healthy but the FEX is not. Orphan ports and peer-gateway behaviour on the parent affect FEX traffic in the same way as any vPC member - see Nexus vPC: Peer-Gateway, Orphan Ports and Health Checks.

Escalation checklist

1. show fex - state and serial, confirms association.
2. show interface fex-fabric - are the expected fabric ports Active and mapped?
3. show logging | include FEX - image mismatch, duplicate FEX ID and version incompatibility appear here.
4. Compare NX-OS release and FEX hardware support before anything else; an unsupported FEX in the fabric produces permanent Offline states.
5. After any cabling change, re-verify host interface to server NIC mapping with Cisco Nexus NX-OS OSPF Configuration with Examples rather than trusting labels.

原文链接：https://www.cisco.com/c/en/us/support/docs/switches/nexus-2000-series-fabric-extenders/200265-Troubleshooting-Fabric-Extender-FEX-Pe.html