NetBackup 5230, 5240 and 5330 appliances using a 10GB PCI network card (x710) may kernel panic during startup

Problem

In certain circumstances a NetBackup appliance using the x710 10GB PCI network card, the NetBackup appliance has been to known to kernel panic when the appliance and operating system starts to perform a hardware probe.

Probing EDD (edd=off to disable)... ok
Starting udev

15-20 seconds after the ‘ Starting udev ‘ message appears the panic occurs.

Note:  If the appliance is booted *without* the cable attached to either the card or switch port, then a successful boot is achieved.   The kernel panic does not occur when the 10GB network ports which are directly attached to the appliance system board are used.
 

Error Message

When the kernel panic occurs, these are the starting lines of the trace.

Call Trace:
[<ffffffffa03dbf58>] ? i40e_write_rx_ctl+0x58/0x90 [i40e]
[<ffffffffa03cb738>] i40e_setup_pf_switch+0x318/0x5a0 [i40
[<ffffffffa03cb738>] i40e_probe+0x1185/0x1be6 [i40e]
 

Cause

Where the x710 10GB PCI network card is connected to switches which have the DCBX (Data Center Bridging Capability Exchange) protocol, which is an extension of Link Layer Data Protocol (LLDP) enabled, during the startup of the appliance the panic occurs.
 
Note:  In this example where the problems were seen the switch make and type was a ‘ Juniper 4500 ‘.
 

Solution

Veritas engineering are aware of the problems with the DCBX protocol ( Etrack 3909672 and 3965629) which affect appliance versions 2.7.3, 3.0, 3.1.x and probably 3.2.   At the time this technical article was written there is no formal solution available, therefore, Veritas engineering suggest for the foreseeable future the DCBX / LLDP protocol should be disabled to help prevent any issues.  Please contact Veritas NetBackup Technical Support quoting Etrack reference 3909672 for further details.
 
There are options to work-around for the problem

• Disable the DCBX / LLDP protocols on the switch.
• Use an alternative switch.
• Disconnect the network from the x710 10GB PCI card during the initial appliance startup (this would need to be done everytime the appliance was booted).  At the point after the hardware has been probed and the appliance continues to install the kernel, then reattach the network.