Nodealias crashes while adding a new node to an existing X440-G2 stack

X440-G2Recently I had to add a new node to an existing stack of 2 X440-G2 switches. The process is quite well documented on the Extreme Networks KB.

I followed this guide step-by-step, and at first everything seemed just fine.

However, only minutes after having completed the procedure, the master node threw an error:

Slot-1 Switch.10 # Process nodealias pid 2033 died with signal 11

Code:
40c708 8ca40000 lw a0,0(a1)
40c70c 00021040 sll v0,v0,0x1
40c710 00821021 addu v0,a0,v0
40c714 <94420000>lhu v0,0(v0)
40c718 3404ffff ori a0,zero,0xffff
40c71c 10440013 beq v0,a0,0x40c76c
40c720 00731821 addu v1,v1,s3
40c724 8c650000 lw a1,0(v1)
40c728 2403004c addiu v1,zero,76

followed by a reboot.

After investigation by GTAC, this seemed to be caused by SNMP querying the node alias interface table while slot3 (the new node) was not completely initialized.

The fix for this bug, EXOS-27469), is fixed in 30.7.1.1-patch-86 and 31.2.1.

So if you are adding a node to an EXOS stack running earlier versions, and you have XMC, watch your timing: XMC SNMP polling interval by default is 5 minutes…

 

Leave a Reply

Your email address will not be published. Required fields are marked *