Contributors: Ravi Kumar Poudapelly, Srikanth Kondapaneni

Introduction

Traditionally on AIX, transceiver statistics were collected and displayed using Advanced Diagnostics. To enter Advanced diagnostics mode, all IP addresses configured on these devices need to be tear-ed down. This causes application outage.

With this feature, device driver would:

A.       Collects and displays transceiver statistics as part of detailed ethernet statistics (collected using command entstat -d entX, where entX is network devices).

B.       Runs check on the collected transceiver statistics to determine the health of transceiver (health check). If any health check fails, corresponding error logs would be recorded. This helps customers and IBM support to isolate transceiver issues early on.

C.      Because “AIX snap” command (used to gather system configuration and statistics) has an existing way to collect detailed ethernet statistics, so, transceiver statistics would also be collected. This helps in saving time required to isolate transceiver issues seen by Customers.

D.      Transceiver statistics are collected when adapter is running in “Dedicated” (non-SRIOV) Mode.

Transceiver Statistics

Now, adapter driver would gather transceiver statistics and displays them as a new section named “QSFP Transceiver Statistics” was added at the end of detailed ethernet statistics output. Following information would be displayed in this section:
           i.    Transceiver Vital Product Data (VPD) Information:

           a.        Vendor Name

           b.        Vendor Part Number

           c.        Vendor Serial Number
           d.        Vendor Organizational Unique Identifier (OUI)

ii.    Transceiver Statistics:

           a.        Wavelength

           b.        Module Speed

           c.        Media Type

           d.        Temperature

           e.        Voltage

           f.         Rx Power

           g.        Tx Power

Detailed explanation for each field:

Today this feature is enabled only for specified adapters:

a)        PCIe4 2-port 100 GbE RoCE x16 adapter (FC EC66 and EC67; CCIN 2CF3)

b)        PCIe4 2-port 100 GbE RoCE x16 adapter (FC EC75 and EC76; CCIN 2CFB)

c)        PCIe5 x16 2-port 200 GbE RoCE adapter (FC EC85 and EC86; CCIN EC2C)

Non-separable cables would not capture wavelength, Rx Power and Tx Power. When adapter is connected with non-separable cables, driver would not display these fields.

As this feature is not supported when adapter is running in SRIOV (Shared) mode, so, transceiver statistics would not be displayed for native VFs, vNICs and HNV.

Sample Transceiver Statistics:

To collect transceiver statistics, detailed ethernet statistics need to be run:

Following is a snippet of transceiver statistics collected for QSFP-28 transceiver module:

Health Check:

Health Check would generate an error when a transceiver parameter falls outside its defined limits.

This function is limited only to display errors to alarm customers; no action would be taken – Driver would continue to send and receive traffic, even with transceiver failure error. This error does not necessarily indicate a transceiver hardware failure. Please contact IBM Customer Support for further assistance and a detailed diagnosis. 

Transceiver statistics collection by the driver is limited to the execution of the entstat command; consequently, health checks occur only when this command is run. So, health check would not be done periodically.

Transceiver health-check errors are recorded with “MLXCENT_TRANSCEIVER_ERR” label and description as “Transceiver Error” in error log. Following is a sample error recorded by transceiver when temperature exceeded specified range (high temperature):

List of different transceiver errors:

Transceiver errors are embedded in error-log as follows:

DEVICE DRIVER INTERNAL STATE

XXXX XXXX XXXX XXXX YYYY YYYY YYYY YYYY ZZZZ ZZZZ ZZZZ ZZZZ

YYYY YYYY YYYY YYYY – This field represents type of transceiver error

Different type of transceiver errors was listed in the table below:

This feature is available to user from AIX 7300-04 and VIOS 4.1.2.0 release onwards.