Recover From a Port Hang (Oracle Solaris)

1. If the interface encounters a soft hang, replumb the device.

   Use the ipadm(1M) command.

2. If the interface encounters a hard hang, reboot the server.

   If the interface encounters another hard hang, try to capture the trace information by using the dtrace(1M) command, as in this example.

   # detrace -F -m 'i40e{trace(timestamp)}'
   > /tmp/dtrace.out
   

3. If the server is panicked, retrieve the crash dump in /var/crash.
4. If the interface encountered a hard hang or a panic, file a CR at My Oracle Support. Attach the last page of the dtrace(1M) output or the crash dump file to the CR.
5. Check for the driver statistics.

   # kstat i40e: 1:statistics
   

6. Use the following parameters for performance tuning in i40e.conf:

   Parameters Description rx_itr The interval of receive interrupts 0 - 4080, 25 (50usec) by default . tx_itr The interval of transmit interrupts 0 - 4080, 25 (50usec) by default. rx_limit_per_intr Maximum number of packet to receive per interrupt16 ~ 4096, 1024 by default. rx_copy_threshold Packet size to determine bcopy or not during receive0 ~ 9216, 128 by default. tx_copy_threshold Packet size to determine bcopy or not during transmit0 ~ 9216, 128 by default

Related Information

• Analyze Why the Device Link Is Missing (Oracle Solaris)

• Analyze Slow Network Performance (Oracle Solaris)

• Analyze Why the Link Is Not Up After Back-To-Back Cable Connection (Oracle Solaris)

• Analyze Why Changing the MTU Does Not Correctly Set the Link Property (Oracle Solaris)