PlantUML DiagramHA Stack
A structured lesson workspace with readable content, hands-on examples, and a clean path to completion.
High availability isn't a single feature — it's a design philosophy applied at every layer. You can have HSRP configured perfectly and still have a single point of failure two racks over that takes down the whole thing. Real HA requires thinking about every component, every link, and every failure mode systematically.
Most engineers only think about one or two layers. The classic mistake: redundant routers with HSRP but a single cable to the internet. The HSRP failover works perfectly, but if the cable fails, both routers lose connectivity. HA requires diversity at every layer.
Redundant hardware in the same rack protects against hardware failure but not power outage or physical damage. Geographically diverse sites protect against both but introduce latency and complexity. Match the HA level to the business requirement — not every system needs five-nines availability.
HSRP/VRRP handle the default gateway for hosts. But the routers still need to know where to forward traffic — that's the routing protocol's job. A complete HA design combines both: HSRP for the host-facing gateway, OSPF for the router-to-router path. If the active HSRP router loses its OSPF path, HSRP tracking detects it and the standby takes over.
Two core routers are configured with HSRP and OSPF. Each connects to a different ISP. Users report that a single ISP outage takes down their internet, even though the other ISP has redundant links. What's the likely cause?
Select the safest next action.
| Failure type | Typical convergence | Notes |
|---|---|---|
| STP loop | 30-50 seconds (classic), <1 sec (RSTP) | Depends on protocol |
| Link failure (routing) | 1-10 seconds with tuned timers | OSPF Hello/Dead tuned |
| FHRP failover | 3-10 seconds | Depends on timers |
| EtherChannel member | <1 second | Remaining links take over |
| Routing protocol restart | 30 seconds to minutes | Full reconvergence |
| Power failure (whole site) | Minutes to hours | Full restore |
Know your convergence budget. If your application can't tolerate more than 5 seconds of downtime, tune timers aggressively — but understand the trade-off in CPU and stability. If you can tolerate 30 seconds, defaults are fine.
HA design is about asking 'what fails next?' after each layer of redundancy. If you've made the router redundant, what about the cable? The power? The ISP? The DNS? Walk down the stack for any system that matters. You'll find the gaps — and then you can decide whether to close them or accept the risk.
Finish the lesson once you have worked through the material. This awards ★ 35 XP.