When using Veritas Cluster Server (VCS), the following error messages from your system logs can indicate a problem with the cluster heartbeat interconnects:

Dec 12 15:26:20 serverb llt: [ID 194859 kern.notice] LLT:10019: delayed hb 935350 ticks from 1 link 0 (qfe:0)
Dec 12 15:26:20 serverb llt: [ID 761530 kern.notice] LLT:10023: lost 18706 hb seq 3448194 from 1 link 0 (qfe:0)
Dec 12 15:26:20 serverb llt: [ID 194859 kern.notice] LLT:10019: delayed hb 935350 ticks from 1 link 1 (eri:0)
Dec 12 15:26:20 serverb llt: [ID 761530 kern.notice] LLT:10023: lost 18706 hb seq 3448194 from 1 link 1 (eri:0)

These types of messages can be seen when you are running two LLT links over the same physical network. This is bad from a design point of view, as it may introduce a single point of failure. However, there are situations where you may have two physical connections into your cluster servers and have the links run over the same VLAN. If you are sure your interconnects are working properly and you are experiencing this error due to the issue described above then you should be able to solve it by changing your /etc/llttab file on all cluster members.

By default, on Solaris, your /etc/llttab file will look something like this:

set-node servera
set-cluster 1
link eri0 /dev/eri:0 - ether - -
link qfe0 /dev/qfe:0 - ether - -
link-lowpri ce0 /dev/ce:0 - ether - -

The second to last field for each of the links is the SAP field, or ethernet type used for the LLT link. This defaults (when specified using -) to 0xCAFE. Two LLT links on the same physical broadcast domain for a cluster cannot share the same SAP ID. If you do this, you may get the above error messages. Assuming this to be your problem (eg, if you run your eri0 and qfe0 links over the same broadcast domain) you can work around the problem by changing your /etc/llttab file to the following:

set-node servera
set-cluster 1
link eri0 /dev/eri:0 - ether 0xCAFE -
link qfe0 /dev/qfe:0 - ether 0xCAFF -
link-lowpri ce0 /dev/ce:0 - ether - -

This tells LLT to use different SAP types for the two links. All cluster members need to have this change made on them and have the cluster node restarted or have llt restarted.