High Availability, Redundancy, and Resiliency
High availability on VSP is mostly about making two chassis look like one, plus protecting the paths that connect them.
SMLT and IST
- Two VSPs form a cluster over an IST (Inter-Switch Trunk). The IST is a special MLT carrying cluster control traffic and must be a direct, dedicated link with a peer IP on its own VLAN.
- SMLT links from access switches, servers or firewalls are dual-homed to both peers; with LACP the access device sees one logical bundle.
- Both peers must configure the same SMLT ID for the same bundle, or traffic is not handled symmetrically.
Supporting mechanisms
- RSMLT extends SMLT to Layer 3: the peer keeps the owner's IP and MAC reachable after a failure, so routing converges without waiting for protocol timers.
- VLACP detects unidirectional or end-to-end link failure faster than the physical layer reports it; enable it on SMLT and IST links where the peer supports it.
- SLPP catches loops introduced by miswiring in an SMLT environment.
- VRRP provides a redundant gateway; VRRP backup-master allows orderly pre-emption.
- Inside the fabric, SPB provides ECMP and fast reconvergence; dual-homed BEBs plus equal-cost IS-IS paths absorb a single link or node loss.
Worked example
Two VSP 8404s serve a rack of servers. MLT 1 carries the IST with ist peer-ip 10.255.255.2 vlan 4000 and ist enable; MLT 2 is the SMLT to the servers with LACP. Verification: show ist shows the peer up and the IST state, show mlt shows both bundles up, and the IST statistics confirm control traffic. If show ist shows the peer down, SMLT still forwards, but the cluster cannot synchronise MAC learning - a common cause of intermittent flooding.
Traps
Running user traffic and the IST over the same link; mismatched IST VLAN or peer IP; mismatched SMLT IDs; assuming SMLT alone gives L3 gateway redundancy (you still need RSMLT or VRRP); enabling STP on SMLT ports; and testing failover without VLACP in place.