Skip to content

OSPF LSA Injection — primer

Help Center primer for the OSPF Link-State Advertisement Injection stress engine. Pairs with ADR 0011 and the ospf-router-peer/ module.

What it tests

NGFW vendors quote BGP scale loudly and OSPF scale rarely — yet OSPF is the protocol they actually run on inside many enterprise DCs. This engine pressures the OSPF subsystem along three orthogonal axes:

  1. Adjacency churn — bring sessions up / down at a controlled rate to measure DR election + Designated Router fail-over time.
  2. LSA flood rate — inject Type-1 (Router), Type-2 (Network) and Type-5 (External) LSAs to fill the LSDB at a chosen rate.
  3. SPF recomputation cost — observe DUT CPU + memory while it re-runs Dijkstra under sustained LSA churn.

Output: T₅₀ / T₉₅ / T₉₉ adjacency-up time, LSDB peak size, SPF runs per minute, CPU envelope, and adjacency-flap recovery.

Three-axis configuration

Axis Options
enabled yes / no
lsa_mix type1-only / type1+type2 / type1+type2+type5 (default)
lsa_rate low (10/s) / medium (100/s) / high (1000/s) / line-rate (>1000/s)

The dashboard pre-validates (SKU, lsa_mix, lsa_rate) against the DUT catalog. A SKU with no published OSPF scale claim returns a warning + a "best-effort capture" recommendation.

Static-route precedence — safety guarantee

Persona forwarding uses static routes (admin distance 1). OSPF-injected routes (admin distance 110) NEVER override them. Persona reachability is unaffected even when the LSDB is fully saturated.

Layered vs standalone

  • Standalone: pick test_kind = bgp-saturation with protocol=ospf modifier (Test Kind #5 covers IGPs as well).
  • Layered: any data-plane test can enable the ospf_layered modifier to run OSPF churn in parallel — observe whether DUT TLS throughput drops when SPF re-runs.

Reading the report

Each OSPF run adds an "Annex L (OSPF)" block:

  • DUT → vendor LSDB capacity claim (if known)
  • Run config → 3 axes
  • Convergence → adjacency-up T₅₀/T₉₅/T₉₉
  • LSDB envelope → peak entry count + memory
  • SPF activity → runs/min + CPU peak during runs
  • Churn signature → adjacency flap p95 recovery

Common patterns

Symptom Likely cause
Adjacency flaps under high LSA rate DUT OSPFd CPU saturated; lower rate or stagger neighbors
SPF runs every second sustained DUT recomputing on every LSA; check vendor SPF dampening config
LSDB capacity reached before rate target DUT undersized for IGP — capture for sales conversation
Persona pods unreachable during run (Should never happen) Verify static-route precedence — admin distance 1 vs 110

Last verified against shipping code: v3.7.0 (2026-05-12).