Spanning tree

From Teknologisk videncenter
Revision as of 10:45, 21 May 2012 by Inckie (talk | contribs) (→‎Root Bridge Election graph)
Jump to: navigation, search

The Spanning Tree protocol are used on Ethernet Switches to avoid broadcast storms.

Spanning Tree has evolved during the years. See Spanning Tree articles

Problems with Switches without Spanning Tree

Broadcast Storms

A broadcast starts when a Ethernet switch receives a broadcast from a Host and there exist a loop. See example below:

Broadcast storm example

Duplicate Ethernet Frames

Another problem with Ethernet loops is duplicate Ethernet frames. In the picture below, switch X can see PC-B mac-address on two ports and send the frame out of each port.

Duplicate Ethernet frame example

Instability in Switch MAC-Address table

Another problem with with Ethernet loops is instability in the Switches MAC-Address table See the picture below and consider:

  • The MAC-Address off PC-B is timed out on both Switches.
  • PC-A send a unicast packet to PC-B's MAC-Address.
  • PC-B's MAC-address is unknown to Switch X which send the Frame out of all ports. (Except the originating port 3)
  • Switch Y receives the Frame to PC-B on port 1 and on port 2.
  • PC-B's MAC-address is unknown to Switch X which send the Frame out of all ports. (Except originating port)
  • Switch X know receives Frames on port 1 and port 2 with source MAC-address o fPC-A
  • Switch X now thinks that PC-A is on Port 1, Port 2 and Port 3.
Broadcast storm example

The Spanning Tree protocol principle

If the Switches has enabled the Spanning Tree Protocol - STP - the Switches discover the loop and close one of the links for traffic. The closed link will be enabled if one of the other links breaks down.

Broadcast storm example

BPDU: Bridge Protocol Data unit

Connected Switches sends BPDU packets to each other, to make a hierarchy among them. The purpose is to build a loop free network.

  • BPDU packets are send as 802.1d multicast packets.
  • Switches not participating in the STP sends the BPDU packets out all ports. (Broadcast)
  • Switches participating in STP receives the packets on the multicast addresses 01-80-C2-00-00-00 and 01-80-C2-00-00-10
  • BPDU packets are send every 2 seconbds.

Selection of Root Bridge

BPDU packets are propagated through the switches, each switch identifying itself with its BID or Bridge Identifier. The switch with the lowest Bridge Identifier will become the Root Switch. Default the Priority Field of the Bridge Identifier is 32768 which will lead to the switch with the lowest MAC address being the Root Switch.

BID or Bridge Identifier

The BID is a Integer consisting of eight Bytes. The first two bytes are priority and the remaining six bytes are the MAC address of the switch.

2 Bytes Priority 6 Bytes MAC Address

Extended System in Bridge ID field

MST extended bridge identifier

MAC address reduction splits the 16-bit field into two fields: a configurable 4-bit field and a nonconfigurable 12-bit field. The 12-bit field carries the VLAN ID (VID) or, with MSTP, the MSTP instance number. The two fields are merged to create the unique Bridge Priority field for a particular VLAN or MSTP instance. The appending MAC address remains the same for all instances. The priority field is default 32768.

Extended Bridge Priority field setting
Priority 4 bits Priority 16 bits Actual MSTP Priority Actual PVST/PVST+ Priority
0 0 0 + MSTP instance 0 + VLAN ID
1 4096 4096 + MSTP instance 4096 + VLAN ID
2 8192 8192 + MSTP instance 8192 + VLAN ID
3 12288 12288 + MSTP instance 12288 + VLAN ID
4 16384 16384 + MSTP instance 16384 + VLAN ID
5 20480 20480 + MSTP instance 20480 + VLAN ID
6 24576 24576 + MSTP instance 24576 + VLAN ID
7 28672 28672 + MSTP instance 28672 + VLAN ID
8 32768 32768 + MSTP instance 32768 + VLAN ID
9 36864 36864 + MSTP instance 36864 + VLAN ID
10 40960 40960 + MSTP instance 40960 + VLAN ID
11 45056 45056 + MSTP instance 45056 + VLAN ID
12 49152 49152 + MSTP instance 49152 + VLAN ID
13 53248 53248 + MSTP instance 53248 + VLAN ID
14 57344 57344 + MSTP instance 57344 + VLAN ID
15 61440 61440 + MSTP instance 61440 + VLAN ID

Root Bridge Election graph

Root bridge election.gif


BPDU Packet

Bridge Protocol Data Unit
Bytes Field name Notes
2 Protocol ID Always 0
1 Version Always 0
1 Message Type Configuration or TCN BPPU
1 Flags
8 Root Bridge ID 2 Bytes priority and 6 Bytes MAC address
4 Path Cost Cost of all links from the transmitting switch to the root bridge

See table below

8 Bridge ID 2 Bytes priority and 6 Bytes MAC address
2 Port ID Transmitting switch port ID
2 Message age in 256's of a second
2 Max age in 256's of a second
2 Hello Time in 256's of a second
2 Forward delay in 256's of a second

BPDU Path Cost

STP Path cost
Link Bandwidth Old STP Cost New STP Cost
4 Mbps 250 250
10 Mbps 100 100
16 Mbps 63 62
45 Mbps 22 39
100 Mbps 10 19
622 Mbps 2 6
1 Gbps 1 4
10 Gbps 0 2

STP Port States

STP Port States
  • Blocking state (max-age 20 sek.)
    • Listening for BPDU’s
    • Data packets are not transmitted or received.
    • The port will not learn MAC addresses.
  • Listening state (forward delay = 15 sek.)
    • Listening for BPDU’s
    • Data packets are not transmitted or received.
    • The port will not learn MAC addresses.
    • All ports remain in this state until the Root Switch is elected.
    • Non-designated ports are blocked.
  • Learning state (forward delay = 15 sek.)
    • Listening for BPDU’s
    • Data packets are received to learn MAC addresses.
    • Data packets are not transmitted.
  • Forwarding state (Normal Operation)
    • Listening for BPDU’s
    • Data packets are received and transmitted.
    • Learning MAC addresses.
  • Disabled state
    • Ports administratively closed, not disabled by STP. (Shutdown)

Enhancements to the Spanning Tree Protocol

Der er adskillige tilføjelser til Ciscos implementering af STP heriblandt

PortFast

PortFast får et Lag 2 LAN interface til at skifte til forwarding tilstand straks. Dette foregår ved at springe over listening og learning og gå direkte til forwarding tilstand.

Anvend kun portfast på porten når den er tilsluttet en end-unit. (PC,Printer...) 

AccessSW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
AccessSW1(config)#<input>int range fastEthernet 0/1 - 24</input>
AccessSW1(config-if-range)#<input>spanning-tree portfast</input>

Eksempel på konfigurering af portfast

BPDU guard

Hvis BPDU guard er konfiguret på en PortFast enablet port, vil porten straks lukke ned hvis der kommer en BPDU ind. 

AccessSW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
AccessSW1(config)#<input>int range fastEthernet 0/1 - 24</input>
AccessSW1(config-if-range)#<input>spanning-tree portfast</input>
AccessSW1(config-if-range)#<input>spanning-tree bpduguard enable</input>

Uplinkfast

Giver tre til fem sekunders konvergens tid - det vil sige før linken er i drift igen - efter en link fejl. Dette foregår ved at springe over listening og learning og gå direkte til forwarding tilstand.

Anvend kun uplinkfast på links mellem switche som ikke er root-switchen.

Uplinkfast kan ligeledes load-balancing mellem redundante lag 2 links ved at anvende uplink groups.

En uplink group er en gruppe af porte der kører spanning tree i PVST - Per-Vlan Spanning Tree - hvor kun en port er i forwarding og resten er blocked. Load-Balancing kan opnås ved at have forskellige porte i forwarding hvor de andre er blocked 

DistSW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
DistSW1(config)#<input>spanning-tree uplinkfast</input>

UplinkFast enables globalt på switchen.

BackboneFast

BackboneFast er en Cisco udvidelse til Spanning-Tree som anvendes i forbindelse med UplinkFast og nedsætter nedetiden på en switch der modtager BPDU'er fra en anden Switch som har mistet forbindelsen til Root-Switchen og selv tror den er Root-Switch. Se eksemplet herunder

1

Normal drift

Netværket er stabilt og fungerer som tiltænkt. Switch-C er i Blocking state, fordi Switch-B har en bedre Bridge-ID end Switch-C. Switch-B er Secondary Root-Bridge.

1: Netværket under normal drift.
2

Linkfejl

Der sker en linkfejl mellem Switch-A og Switch-B. Switch-B har nu mistet sin Root-port og tror nu at den er Root-Bridge, og begynder nu at sende sine egne BPDU pakker ud af Fa0/2.

Switch-C modtager nu BPDU-pakker fra Switch-A og Switch-B. Pakkerne fra Switch-B har en dårligere Bridge-ID end dem fra Switch-A.

BPDU-pakker sendes og modtages også igennem blockede porte.

2: Linkfejl mellem Switch-A og Switch-B.
3

Normal Spanning-Tree uden BackboneFast

Under normal Spanning-Tree vil Switch-C ignorere BPDU-pakkerne fra Switch-B indtil Max-Age timeren udløber. (normalt 20 sekunder) Hvis den stadig modtager BPDU-pakker fra Switch-B vil Switch-C begynde at sende BPDU-pakkerne fra Switch-A ud på Fa0/2 til Switch-C. Fa0/2 vil herefter skifte til Listening tilstand (15 sekunder) og Learning tilstand (15 sekunder).

I alt tager det Max-Age + Listening + Learning sekunder før netværket er i drift igen. 20 + 15 + 15 = 50 sekunder.

2: Linkfejl mellem Switch-A og Switch-B.
4

Spanning-Tree med BackboneFast

Når backboneFast er enablet på alle switche, vil Switch-C reagere anderledes end beskrevet ovenfor når den modtager BPDU pakker fra Switch B.

Når Switch-C modtager en BPDU-pakke fra Switch-B tyder det på at Switch-B har tabt kontakten med Root-Brigden(Switch-A). Det kunne også betyde at Switch-C måske har mistet kontakten til Root-Brigden(Switch-A). Derfor vil Switch-C udsende RLQ (Root-Link Query) ud på alle porte som ikke er Designated-Ports(Upstream). I eksemplet udsender Switch-C RLQ pakker på port Fa0/1 til Switch-A.

Switch-A - som er Root-Bridge og har enablet BackboneFast - sender Switch-A en RLQ response pakke tilbage til Switch-C. Denne pakke fortæller Switch-C at den stadig har kontakt med Root-Bridgen. Derfor behøver Switch-C ikke vente i Max-Age (20 sekunder) men kun at sende fa0/2 igennem Listening (15 sekunder) og Learning (15 sekunder) tilstandende. Nedetiden er hermed nedbragt fra 50 sekunder til 30 sekunder.

2: Linkfejl mellem Switch-A og Switch-B.
DistSW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
DistSW1(config)#<input>spanning-tree uplinkfast</input>
DistSW1(config)#<input>spanning-tree backbonefast</input>

Loop Guard

Loop-Guard forhindrer at en Alternate-Port eller Root-Port bliver en Designated-Port hvis BPDU'erne udebliver. Dette kan forhindre en Loop - og dermed en Broadcast-storm hvis der skulle opstå en unidirectional link. Se også UDLD herunder. 

Dist-SW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
Dist-SW1(config)#<input>interface gigabitEthernet 0/1</input>
Dist-SW1(config-if)#<input>spanning-tree guard loop</input>

Root Guard

Root-Guard kan anvendes til at afvise en Switch som har en bedre Bridge-ID end den tiltænkte Root-Switch. Root-Guard tvinger en port til at være en Designated-port. Altså en port der vender væk fra Root-Brigden. Hvis der kommer en Switch med en bedre Bridge-ID end den rigtige Root-Switch vil porten lukke ned, og derved stoppe BPDU pakkerne fra den falske Root-Switch. 

Dist-SW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
Dist-SW1(config)#<input>interface gigabitEthernet 0/1</input>
Dist-SW1(config-if)#<input>spanning-tree guard root</input>
Root guard.png

UDLD - Unidirectional Link Detection

UDLD overvåger den fysiske konfiguration af portene og opdager hvis en port har en en-vejs forbindelse - Unidirectional - og lukker porten ned.

Global enabling af UDLD 

Dist-SW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
Dist-SW1(config)#<input>udld enable</input>

Enabling af UDLD på interface 

Dist-SW1#<input>configure terminal</input>
Enter configuration commands, one per line.  End with CNTL/Z.
Dist-SW1(config)#<input>interface gigabitEthernet 0/1</input>
Dist-SW1(config-if)#<input>udld enable</input>

MST: Multiple Spanning Tree protocol

Ved MST kan man definere flere Spanning Tree instancer. I eksemplet nedenunder er der defineret to MST instancer. Instans 1 bygger et spanning-tree for VLAN 10 til 50 og instans 2 bygger et spanning-tree for VLAN 51 til 99. For at load-balance mellem forskellige switche og trunks og for at få redundans er Switch 1 standard root-switch for instans 1 og Switch 2 standard root-switch for instans 2. 

hostname Switch-1
spanning-tree mode mst
spanning-tree extend system-id
!
spanning-tree mst configuration
 name TEKKOM
 revision 1
 instance 1 vlan 10-50
 instance 2 vlan 51-99
!
spanning-tree mst 1 priority 24576
spanning-tree mst 2 priority 32768

!
vlan internal allocation policy ascending
hostname Switch-2
spanning-tree mode mst
spanning-tree extend system-id
!
spanning-tree mst configuration
 name TEKKOM
 revision 1
 instance 1 vlan 10-50
 instance 2 vlan 51-99
!
spanning-tree mst 1 priority 32768
spanning-tree mst 2 priority 24576

!
vlan internal allocation policy ascending
Retrieved from "http://mars.merhot.dk/w/index.php?title=Spanning_tree&oldid=22743"