cmaven

Worknotes

:bulb: When a 25G link won’t come up, this is how to test the GBIC (SFP28 transceiver) modules in the switch, diagnose the cause, and reach a conclusion.
Topology: Server (25G NIC) ── fiber ── Arista switch (SFP28 port)
Key diagnostic points: Tx/Rx optical power, Bias Current, errdisabled reason, inventory module identification, server-side ethtool cross-check

The CLI output in this post is drawn from real work testing 25G SFP28 transceivers on an actual device (DCS-7050SX3-48YC8C-F / EOS 4.30.4M, hostname sw-core-01). Device identifiers such as serial numbers and MAC addresses are masked (xxxx), and IPs use documentation example values. Substitute your own values when diagnosing in practice.

[00] Goal and Setup

A new 25G NIC is connected to an Arista switch but the link won’t come up. The goal is to determine whether the cause lies in the transceiver (GBIC). The test subjects are 25G SFP28 optics across several ports.

Item Value
Switch sw-core-01 (DCS-7050SX3-48YC8C-F / EOS 4.30.4M)
Test ports Ethernet21, Ethernet23, Ethernet35, Ethernet37
Transceivers OEM 25GBASE-SR, MikroTik XS+31LC10D (25GBASE-LR)
Server NIC ens801f0np0 (25G SFP28)

Looking at one or two ports invites the wrong conclusion (“just that module is bad”). Cross-test several ports with different modules to separate an individual-module fault from a policy issue.

[01] Test Method — What to Check, With Which Commands

GBIC diagnosis checks four things in order.

Step What Command
1 Port status / transceiver type show interfaces EthernetN status
2 Optical power (Tx/Rx) and Bias show interfaces EthernetN transceiver
3 errdisabled reason show interfaces status errdisabled
4 Module vendor/model/serial show inventory

The crux is the three optical-power values.

  • Bias Current (mA) — transmit laser drive current. 0.00 mA means the transmitter is off.
  • Tx Power (dBm) — optical output the switch emits. -30.00 dBm means effectively no transmission.
  • Rx Power (dBm) — incoming optical signal from the far end (server). A normal range means the cable and the far-end transmitter are alive.

[02] Collecting Per-Port Transceiver State

[02-1] Ethernet21 — Port Status



1
2
3



sw-core-01#show interfaces Ethernet21 status
Port       Name   Status       Vlan     Duplex Speed  Type         Flags Encapsulation
Et21              errdisabled  1        full   25G    25GBASE-SR

The transceiver type is correctly detected as 25GBASE-SR at 25G. But the status is errdisabled — not a cable-out notconnect, but the switch forcibly disabling the port after detecting a specific error condition.

[02-2] Ethernet21 — Optical Power



1
2
3
4
5
6



sw-core-01#show interfaces Ethernet21 transceiver
                               Bias      Optical   Optical
          Temp       Voltage   Current   Tx Power  Rx Power
Port      (Celsius)  (Volts)   (mA)      (dBm)     (dBm)     Last Update
-----     ---------  --------  --------  --------  --------  -------------------
Et21       31.11      3.31      0.00     -30.00    -0.28     0:00:01 ago
Item Value Meaning
Bias Current 0.00 mA No transmit-laser drive current
Tx Power -30.00 dBm Switch optical output is effectively nil
Rx Power -0.28 dBm Optical signal from the server is detected

So the switch receives but does not transmit. This asymmetry is the key clue.

[02-3] Ethernet21 — Detailed Counters



1
2
3
4
5
6
7
8



sw-core-01#show interfaces Ethernet21
Ethernet21 is down, line protocol is down (errdisabled)
  Hardware is Ethernet, address is xxxx.xxxx.5f06 (bia xxxx.xxxx.5f06)
  Ethernet MTU 9214 bytes, BW 25000000 kbit
  Full-duplex, 25Gb/s, auto negotiation: off, uni-link: n/a
  ...
     0 input errors, 0 CRC, 0 alignment, 0 symbol, 0 input discards
     0 output errors, 0 collisions

CRC, symbol, and alignment errors are all 0. This is not errors accumulating after a link came up — the port was blocked before reaching normal operation.

[02-4] Ethernet23 — Same Pattern (25GBASE-SR)



1
2
3



sw-core-01#show interfaces Ethernet23 transceiver
Port      ...  Current(mA)  Tx Power(dBm)  Rx Power(dBm)  Last Update
Et23      ...  0.00         -30.00         -0.38          0:00:01 ago

Et23 matches Et21 exactly: Bias 0.00 mA / Tx -30.00 dBm / Rx -0.38 dBm. Two ports with identical symptoms point away from an individual port fault and toward a module/policy issue.

[03] Cross-Test — Does It Reproduce on Other Modules?

To reach a sound conclusion, check whether different transceivers show the same symptom.

[03-1] Ethernet35 — A Different OEM SR Module



1
2



sw-core-01#show interfaces Ethernet35 transceiver
Et35      ...  0.00(mA)   -30.00(Tx dBm)   -0.74(Rx dBm)

transceiver detail makes it clearer against the thresholds.



1
2
3
4
5
6
7



sw-core-01#show interfaces Ethernet35 transceiver detail
           Tx Power      Low Alarm   Low Warn
Port       (dBm)         (dBm)       (dBm)
-------    ------------  ----------  ----------
Et35       -30.00        -11.40      -8.40
           Rx Power
Et35       -0.75         (normal range)

The current Tx Power -30.00 dBm is far below the -11.40 dBm Low Alarm, while Rx Power is in normal range → the switch is not driving the transmit laser.

[03-2] Ethernet37 — A MikroTik 25GBASE-LR Module

Test an LR module, not just the SR family.



1
2
3
4
5
6
7
8
9
10



sw-core-01#show interfaces Ethernet37 status
Et37              errdisabled  1        full   25G    25GBASE-LR

sw-core-01#show interfaces Ethernet37 transceiver detail
           Current   Low Alarm
Et37       0.00 mA   1.00 mA
           Tx Power  Low Alarm
Et37       -30.00    -7.00 dBm
           Rx Power  Low Alarm
Et37       1.05      -19.00 dBm   (normal range)

The 25GBASE-LR MikroTik XS+31LC10D also shows Bias 0.00 mA / Tx -30.00 dBm with only Rx normal. The fact that SR and LR alike show the same symptom is decisive.

[04] Confirming the errdisabled Reason — The Clinching Evidence



1
2
3
4
5
6
7
8



sw-core-01#show interfaces status errdisabled
   Port       Name             Status         Reason
---------- ---------------- ----------------- ----------------
   Et21                        errdisabled    xcvr-unsupported
   Et23                        errdisabled    xcvr-unsupported
   Et35                        errdisabled    xcvr-unsupported
   Et37                        errdisabled    xcvr-unsupported
   ... (Et1, Et3, ... many more ports the same)

All four ports — and many others — show the reason xcvr-unsupported. The switch classifies the installed transceivers as “unsupported” and blocks the ports.

xcvr-unsupported typically arises from:

  • An SFP28 module not on Arista’s certified/compatible list
  • Optics coded exclusively for another vendor
  • A module whose EEPROM data doesn’t match Arista equipment
  • An EOS policy that blocks unsupported transceivers

[05] Identifying the Modules via inventory

Use show inventory to see exactly what’s blocked.



1
2
3
4
5
6
7
8



sw-core-01#show inventory
System has 56 switched transceiver slots
  Port Manufacturer     Model            Serial Number    Rev
  ---- ---------------- ---------------- ---------------- ----
  21   OEM              25GBASE-SR       PS3Hxxxxxx       0002
  23   OEM              25GBASE-SR       PS3Hxxxxxx       0002
  35   OEM              25GBASE-SR       PS3Hxxxxxx       0002
  37   MikroTik         XS+31LC10D       OLxxxxxxxxxx     1.0

The key point: the switch reads the vendor/model/serial/revision correctly.



1



37   MikroTik   XS+31LC10D   OLxxxxxxxxxx   1.0

So it’s not a case of failing to read the EEPROM. It reads the data but does not recognize the module as Arista-compatible, and blocks it as xcvr-unsupported.

[06] Server-Side Cross-Check (ethtool)

Confirm the switch-side diagnosis once more from the server.



1
2
3
4
5
6



root@server-a:~# ethtool ens801f0np0
        Supported link modes:   25000baseSR/Full ...
        Supported FEC modes: None        RS      BASER
        Speed: Unknown!
        Duplex: Unknown! (255)
        Link detected: no (No cable)

The server NIC supports 25G SR (25000baseSR/Full) and RS-FEC, yet reports Link detected: no (No cable).

Since the switch has its transmitter off (Tx Power -30.00 dBm / Bias 0.00 mA), the server seeing “No cable” is the natural consequence — not a server NIC capability problem.



1
2
3



root@server-a:~# ethtool --show-fec ens801f0np0
Configured FEC encodings: Auto
Active FEC encoding: None

Active FEC: None is merely a result of the link being down, not evidence of an FEC mismatch.

[07] Interpreting the Physical Link State

Putting the data together, the optical path looks like this:



1
2



Server NIC Tx  ─────────────▶  Switch Rx detected (Rx Power normal)
Server NIC Rx  ◀────────────X  Switch Tx disabled (Bias 0mA, Tx -30dBm)
Evidence Value
Switch Rx Power Et21 -0.28 / Et23 -0.38 / Et35 -0.74 / Et37 1.05 dBm (detected)
Switch Tx Power all four ports -30.00 dBm (no transmit)
Switch Bias all four ports 0.00 mA (laser not driven)
errdisabled reason all four ports xcvr-unsupported
Server Link detected: no (No cable)

[08] Root-Cause Conclusion

Priority Candidate cause Assessment
1 Arista transceiver compatibility/support policy Very high
2 OEM/third-party 25G SR·LR modules classified unsupported by EOS Very high
3 unsupported-transceiver unlock/license not applied High
4 Fiber cable problem Low (Rx signal detected)
5 Server NIC configuration Low (25G·RS-FEC supported)
6 FEC mismatch Secondary (link block precedes it)
7 MTU/IP settings Post-link-up check

Conclusion: before cable type (SR/LR) or server settings, the issue is that the Arista switch blocks the installed third-party/OEM 25G SFP28 transceivers as unsupported. It’s not a single bad module (different SR and LR modules all show the same symptom).

[09] Recommended Actions

The most stable fix. Capture module info beforehand to match against the compatibility list.



1
2



show interfaces Ethernet21 transceiver detail
show interfaces transceiver detail

Check: Vendor name, Vendor PN, Vendor SN, Identifier, Connector, Nominal bitrate.

[09-2] Allow Unsupported Transceivers for Testing

If this is a test environment and policy permits, consider:



1
2
3



configure terminal
service unsupported-transceiver
end

Then re-enable the port:



1
2
3
4
5



configure terminal
interface Ethernet21
   shutdown
   no shutdown
end

:warning: On this device (EOS 4.30.4M / 7050SX3), service unsupported-transceiver once returned % Incomplete command, so an additional argument or an unlock/license key may be required. Rather than guessing, confirm with your vendor/Arista. In production this may void vendor support, so use it only for validation/temporary purposes.

Once the transceiver issue is resolved, align RS-FEC on both sides for 25G stability.



1
2
3
4
5



# Switch
interface Ethernet21
   speed forced 25gfull
   error-correction encoding reed-solomon
   no shutdown





1
2



# Server
sudo ethtool --set-fec ens801f0np0 encoding rs

[10] Post-Recovery Verification

Step What Command Expected
1 errdisabled cleared show interfaces status errdisabled Port no longer listed
2 Tx optical output restored show interfaces Ethernet21 transceiver Tx Power not -30 dBm, Bias not 0 mA
3 Server link ethtool ens801f0np0 Speed: 25000Mb/s, Link detected: yes
4 FEC ethtool --show-fec ens801f0np0 Active FEC encoding: RS
5 IP connectivity ping -c 5 10.0.0.2 0% packet loss
6 Throughput iperf3 -c 10.0.0.2 -P 8 -t 30 ~23 Gbps or more

[11] Conclusion

Following Et21·Et23 (OEM 25GBASE-SR), the cross-tests on Et35 (OEM SR) and Et37 (MikroTik XS+31LC10D 25GBASE-LR) all showed errdisabled / xcvr-unsupported.



1
2
3



On Arista DCS-7050SX3-48YC8C-F / EOS 4.30.4M,
the installed OEM 25GBASE-SR and MikroTik 25GBASE-LR transceivers
are all classified as unsupported, leaving the ports errdisabled.

Because the switch ports emit no transmit light (Bias 0 mA / Tx -30 dBm / Rx detected), the server sees “No cable” — but the actual cause is the switch blocking unsupported transceivers. The first action is either (1) replacing with Arista-compatible 25G SFP28 (SR/LR) modules, or (2) obtaining the unsupported-transceiver unlock/license via your vendor/Arista — after which you validate FEC, cabling, MTU, IP, and iperf3 throughput.

You may also enjoy