pa8mm.net

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== STP ======

==== Algorithm ====

How it works:
  - Each switch generates an ID for itself (PBDU ID)
    - BPDU ID has two parts - //priority:System-Mac//
  - Each switch sends BPDUs out of all its ports (as multicast)
  - After BPDUs are exchanged, **Root** switch is elected
    - Root is having the smallest //priority//, and if equal - smallest //System-Mac//
    - Only Root switch keeps sending BPDUs
    - On the Root switch all ports are //Designated//
  - All non-Root switches will elect a **Root** port - port closest to the Root switch
    - Root port is having the smallest **Cost** to the Root switch (cost is based on the speed)
    - If there are parallel links, port with the smallest number wins
  - On all non-Root switches, elect //Designated// port and define //Blocked// ports
    - //Designated// port is the port closest (cost-wise) to the Root
    - If cost is the same - the port on the switch with lowest System-MAC becomes a //Designated// port


==== Port States ====

  - Disabled
  - Listening - switch sends and receives BPDUs. It lasts 15 seconds and it's where Root is elected
  - Learning - switch is building MAC address table. Also lasts 15 seconds 
  - Forwarding
    - Root Port
    - Designated Port

==== During Failure ====

Root port sends BPDUs every 2 seconds. If non-root switch stops receiving BPDUs over Root port for 20 seconds, it starts the //Listening// and //Learning// process on the //Blocked// port. Failover time is 50 seconds.

==== RSTP ====

RST improves convergence time. Instead of having a //Blocked// port and spending too much time in //Listening// and //Learning//, we have **Alternate** port. Basically it's a second best port if Root port fails. When Root port fails, switch waits **max time out** (20 seconds) and starts forwarding traffic over the new Root port

==== PORT-FAST ====

When port flaps, switch sends TCN (topology change notification) message, which forces all switches to re-calculate the root switch. This is not very efficient when port towards the client flaps. In order to mitigate that, customer facing port is configured with **spanning-tree portfast**. This port will be always on (forwarding). 

==== BPDU Guard ====

We should not receive BPDUs on the customer facing ports. There is a feature which sets the port to **err-disabled** state if switch receives a BPDU on that port.

==== Root Guard ====

Configuration on the Root switch to protect hijacking the Root role. Configured on the ports.

==== MSTP ====

Instead of sending BPDU for each VLAN, MSTP aggregates VLANs into the instances and send a single BPDU for that group.

On Arista, by default MST instance 0 is created and all VLANs belong to that instance.