How do you protect a fabric from losing its subnet manager?
Why A is the answer
Run a standby SM
A standby takes over.
In detail
- A standby takes over if the master fails.
- Run it on a separate host.
Use case
An illustrative scenario, not a real incident.
A master SM fails, and a standby takes over.
See it in different ways
flowchart LR A["Discover fabric"] --> B["Assign LIDs"] B --> C["Program routing tables"] C --> D["Ports go Active"] D --> E["Standby SM takes over on failure"]
sequenceDiagram participant M as Master SM participant S as Standby SM participant F as Fabric M->>F: Configures the fabric S->>M: Checks that it is alive S->>F: Takes over if the master fails
stateDiagram-v2 state "Master SM" as s0 state "Master fails" as s1 state "Standby becomes master" as s2 [*] --> s0 s0 --> s1: failure s1 --> s2 s2 --> [*]
Key point: Run a standby SM.
Why the other options are wrong
- B. Reboot the master each week
- A reboot is not protection.
- C. Use only one switch
- One switch is a single point of failure.
- D. Disable the SM after startup
- The SM must keep running.