Physical Links

The physical cables or fibers that connect the switches to each other and to servers can be monitored with the following commands:

Transceivers

core@control-1 ~ $ kubectl fabric inspect switch --name ds5000-02 --transceivers
NAME         PROFILE             ROLE     GROUPS    SERIAL                    STATE    GEN      APPLIED           HEARTBEAT
ds5000-02    Celestica DS5000    spine              R4028F2B125421GD200304    Ready    55/55    20 minutes ago    6 seconds ago

Transceivers:
NAME     OPER      DESCRIPTION                          CLASS    CONNECTOR       VENDOR     PART                SERIAL           CMIS
E1/1     active    OSFP112 800GBASE-DR8                 FIBER    MPO_2x12        T1NEXUS    T1O112-800G-2DR4    T100AC006346     Ready 4.0 (4)
E1/5     active    OSFP112 2x(400GBASE-CR4-DAC)-2.0M    DAC      NO_SEPARABLE    FS         OSFP-800G-2QPC02    C2504186220      Ready 5.0 (0)
E1/61    active    OSFP112 2x(400GBASE-CR4-DAC)-2.0M    DAC      NO_SEPARABLE    FS         OSFP-800G2OPC015    C2504422334      Ready 5.0 (0)
E1/62    active    OSFP112 2x(400GBASE-CR2-DAC)-1.0M    DAC      NO_SEPARABLE    FS         OSFP-800G-PC01      C2504140951-1    Ready 4.0 (0)
E1/63    active    OSFP112 2x(400GBASE-CR2-DAC)-1.0M    DAC      NO_SEPARABLE    FS         OSFP-800G-PC01      C2504422332-2    Ready 4.0 (0)
E1/64    active    OSFP112 2x(400GBASE-CR2-DAC)-1.0M    DAC      NO_SEPARABLE    FS         OSFP-800G-PC01      C2504422333-1    Ready 4.0 (0)

Use flags for more details: -d/--details (e.g. firmware), -p/--ports, -t/--transceivers, -c/--counters, -l/--lasers

Lasers

The status of the lasers is also available through the inspect commands:

core@control-1 ~ $ kubectl fabric inspect switch --name ds5000-02 --lasers
NAME         PROFILE             ROLE     GROUPS    SERIAL                    STATE    GEN      APPLIED           HEARTBEAT
ds5000-02    Celestica DS5000    spine              R4028F2B125421GD200304    Ready    55/55    21 minutes ago    11 seconds ago

Laser Status:
NAME    CHANNELS IN / OUT ( BIAS )
E1/1    0: 2.61/1.88 dBm (36.56 mA)    1: 1.07/1.72 dBm (36.56 mA)    2: 1.72/1.68 dBm (36.56 mA)    3: 2.38/1.68 dBm (36.56 mA)

Use flags for more details: -d/--details (e.g. firmware), -p/--ports, -t/--transceivers, -c/--counters, -l/--lasers

Often a value of -40dBm on the transmit or receive side indicates that the laser is not on. If supported, check the CMIS status of the optic.

Diagnosing a Down Fiber Link

When kubectl fabric inspect lldp reports an expected neighbor missing on a port, or kubectl fabric inspect bgp reports an underlay neighbor that never reaches Established, the most common root cause is a layer 1 fault (optic, fiber cable, or connector) on that port. The steps below narrow the fault to a single switch port using kubectl fabric inspect.

1. Identify the affected port

kubectl fabric inspect lldp prints a per-switch table of the LLDP neighbors it observes against what the wiring expects. A port with a blank NEIGHBOR / PORT / DESCRIPTION is missing its expected neighbor entirely:

core@control-1 ~ $ kubectl fabric inspect lldp
Switch: ds5000-01 (actual←→expected)
PORT       CONNECTION                      TYPE        NEIGHBOR       PORT            DESCRIPTION
E1/21/1    ds5000-01--mesh--ds5000-02      mesh
E1/22/1    ds5000-01--mesh--ds5000-02      mesh        ds5000-02      E1/22/1         Hedgehog Fabric
...
Switch: ds5000-02 (actual←→expected)
PORT       CONNECTION                        TYPE         NEIGHBOR       PORT                   DESCRIPTION
E1/21/1    ds5000-01--mesh--ds5000-02        mesh
E1/22/1    ds5000-01--mesh--ds5000-02        mesh         ds5000-01      E1/22/1                Hedgehog Fabric
...

Both sides of the link report an empty neighbor on E1/21/1 while the sibling port E1/22/1 on the same connection reports its peer normally. The fault is on that specific physical link between ds5000-01 and ds5000-02, not on either switch as a whole. A working sibling on the same connection rules out a configuration problem shared by the two ports.

kubectl fabric inspect bgp shows the same issue from the routing perspective. The row for the affected port reports STATE as active (a BGP peer that is trying to connect but never reaches established) while other fabric peers on the same switch are established:

core@control-1 ~ $ kubectl fabric inspect bgp
Switch: ds5000-01
TYPE      PORT       VRF        NEIGHBOR        REMOTE NAME          CONNECTION                    STATE          LAST ESTABLISHED
fabric    E1/21/1    default    172.30.128.1    ds5000-02/E1/21/1    ds5000-01--mesh--ds5000-02    active         a long while ago
fabric    E1/22/1    default    172.30.128.3    ds5000-02/E1/22/1    ds5000-01--mesh--ds5000-02    established    a long while ago
...

Both outputs name the same port (E1/21/1), so either starting symptom leads to the same link.

2. Confirm the port is admin up but operationally down

core@control-1 ~ $ kubectl fabric inspect switch --name ds5000-01 --ports
NAME         PROFILE             ROLE           GROUPS    SERIAL                    STATE    GEN    APPLIED          HEARTBEAT
ds5000-01    Celestica DS5000    server-leaf              R4028F2B094A15GD200125    Ready    3/3    6 minutes ago    10 seconds ago

Ports:
NAME       NOS            TYPE        CONNECTION / MODE               ADM / OP     SPEED        TRANSCEIVER                          TRANSC / CMIS
E1/21      Ethernet160                                                                          OSFP112 800GBASE-DR8                 active/ready
E1/21/1    Ethernet160    mesh        ds5000-01--mesh--ds5000-02      up/down      800G
E1/22      Ethernet168                                                                          OSFP112 2x(400GBASE-CR2-DAC)-1.0M    active/ready
E1/22/1    Ethernet168    mesh        ds5000-01--mesh--ds5000-02      up/up        800G

Note

The NAME column is the fabric API port name used in wiring objects and other kubectl fabric inspect commands (for example E1/21/1). The NOS column is the underlying NOS interface name (for example Ethernet160). Both refer to the same port.

Locate the affected port in the ADM / OP column. up/down means the switch is trying to bring the port up but the PHY is not training, which points at the optical path (cable, connector, or transceiver). down/down means the port is intentionally disabled and the fault is at the configuration layer instead; stop here and review the wiring object.

In the example above, E1/21/1 is up/down while the sibling port E1/22/1 on the same connection is up/up, which confirms the fault is specific to one cable rather than a switch-wide configuration issue.

3. Read the per-lane laser powers

core@control-1 ~ $ kubectl fabric inspect switch --name ds5000-01 --lasers
NAME         PROFILE             ROLE           GROUPS    SERIAL                    STATE    GEN    APPLIED          HEARTBEAT
ds5000-01    Celestica DS5000    server-leaf              R4028F2B094A15GD200125    Ready    3/3    6 minutes ago    4 seconds ago

Laser Status:
NAME     CHANNELS IN / OUT ( BIAS )
E1/2     0: -0.47/2.83 dBm (28.25 mA)     1: 1.47/2.82 dBm (28.25 mA)      2: 0.66/2.54 dBm (28.25 mA)      3: 0.45/1.96 dBm (28.25 mA)
E1/17    0: -40.00/2.33 dBm (83.01 mA)    1: -40.00/2.24 dBm (83.10 mA)    2: -40.00/2.27 dBm (83.23 mA)    3: -40.00/2.28 dBm (83.13 mA)
E1/21    0: -40.00/0.32 dBm (28.02 mA)    1: -40.00/0.48 dBm (28.02 mA)    2: -40.00/0.45 dBm (28.02 mA)    3: -40.00/0.51 dBm (28.02 mA)
E1/66    0: -2.16/-0.93 dBm (36.61 mA)

Each row is one port. IN is receive optical power in dBm, OUT is transmit optical power in dBm, and BIAS is transmit laser bias current in mA. Run the same command on the far end of the affected link (ds5000-02 in this example) and compare.

4. Interpret the pattern

Using the example above, port E1/2 is healthy (all IN and OUT values within a few dBm of zero). Port E1/21, the one named in the original LLDP/BGP error, shows all IN values at -40.00 dBm while OUT values are normal: the transceiver is lasing, but no light comes back from the far end. Port E1/17 shows the same pattern but belongs to an unused port and can be ignored here.

Common patterns on both ends and what they indicate:

• All IN values near -40 dBm, all OUT values normal: the patch cord is unplugged, cut, or inserted with the wrong polarity. The far end is not sending light back on any lane.
• Some lanes at -40 dBm, others near 0 dBm, symmetric on both ends: one group of fibers inside the patch cord is damaged, dirty, or miswired; the remaining fibers are fine.
• All OUT values at -40 dBm on one end only: the transmit lasers on that transceiver are not emitting. Suspect a failed or partially inserted optic on that end.
• IN values swing between healthy and below -10 dBm: a dirty end face or bent fiber. Clean and reseat before replacing hardware.

The -40 dBm value is the "no light" measurement floor the module reports when the receiver sees essentially no input. The low-alarm threshold for most modern optics is around -10 dBm; any lane below that threshold is enough to keep the PHY from training, even when the other lanes are healthy.

Ports

The inspect command will also show the connections and counters on a specific port:

core@control-1 ~ $ kubectl fabric inspect switchport -n spine-02/E1/1
Used in Connection spine-02--fabric--leaf-01:
fabric:
  links:
  - leaf:
      ip: 172.30.128.21/31
      port: leaf-01/E1/10
    spine:
      ip: 172.30.128.20/31
      port: spine-02/E1/1
  - leaf:
      ip: 172.30.128.23/31
      port: leaf-01/E1/11
    spine:
      ip: 172.30.128.22/31
      port: spine-02/E1/2

Port Counters (↓ In ↑ Out):
SPEED    UTIL  %         BITS / SEC IN    BITS / SEC OUT    PKTS / SEC IN    PKTS / SEC OUT    CLEAR    ERRORS      DISCARDS
25G      ↓   0 ↑   0     ↓ 2,432          ↑ 2,224           ↓ 3              ↑ 3               -        ↓ 0 ↑ 0     ↓ 2 ↑ 0