Sandeep Agarwal's blog

Recently we got an escalation from the field on AIX in which we saw that gab paniced the node because it got a packet with a wrong sequence number. On further debugging we found that everything was correct except that 1 bit in the gab sequence number was flipped from 0 to 1 thus causing gab to panic the node since it suspected data corruption.
 

If you want to reduce extra traffic on the LLT links one small thing to do would be to add the following line to /etc/llttab on all the nodes in your cluster (say 2 node cluster):
exclude 2-31

It seems that some customers want to create a trunk out of two links and then create vlans over that trunk and then run LLT over those two vlans.

This is how to do this on Solaris:

Create a trunk (aggregation first) out of the two bge links:
dladm create-aggr -d bge2 -d bge3 1 <--- to create a aggr1

Create vlans out of aggr1:

If you have LLT working over the ce driver on Solaris, one small change that you can make to help performance problems in your cluster would be the following:

The /etc/system file on all hosts should have:
set ce:ce_no_tx_lb=1
set ce:ce_taskq_disable=1
set ce:ce_reclaim_pending=1

A reboot is required.