First of all let's explain why we should use Link Dampening?
Interface state changes occur when interfaces are administratively brought up or down or if an interface state changes. Every time an interface changes a state or flaps, routing protocols are notified of the status of the routes that are affected by the change in state. These protocols go through the momentous task of re-converging. Flapping; therefore, puts the status of entire network at risk of transient loops and black holes. Link dampening minimizes the risk created by flapping by imposing a penalty for each interface flap and decaying the penalty exponentially. After the penalty exceeds a certain threshold, the interface is put in an Error-Disabled state and for all practical purposes of routing, the interface is deemed to be “down.” After the interface becomes stable and the penalty decays below a certain threshold, the interface comes up again and the routing protocols re-converge.
Interface state changes occur when interfaces are administratively brought up or down or if an interface state changes. Every time an interface changes a state or flaps, routing protocols are notified of the status of the routes that are affected by the change in state. These protocols go through the momentous task of re-converging. Flapping; therefore, puts the status of entire network at risk of transient loops and black holes. Link dampening minimizes the risk created by flapping by imposing a penalty for each interface flap and decaying the penalty exponentially. After the penalty exceeds a certain threshold, the interface is put in an Error-Disabled state and for all practical purposes of routing, the interface is deemed to be “down.” After the interface becomes stable and the penalty decays below a certain threshold, the interface comes up again and the routing protocols re-converge.
Dampening parameters:
Syntax: dampening [[[[half-life]
[reuse-threshold]] [suppress-threshold]] [max-suppress-time]]
·
half-life
o The number
of seconds after which the penalty is decreased. The penalty decreases half
after the half-life period expires. The range is from 1 to 30 seconds. The
default is 5 seconds.
·
reuse-threshold
o The number
as the reuse threshold, the penalty value below which the interface state is
changed to “up”. The range is from 1 to 20000. The default is 750.
·
suppress-threshold
o The number
as the suppress threshold, the penalty value above which the interface state is
changed to “error disabled”. The range is from 1 to 20000. The default is 2500.
·
max-suppress-time
o The maximum
number for which a route can be suppressed. The default is four times the
half-life value. The range is from 1 to 86400. The default is 20 seconds.
Dampening algorithm:
With each flap, Dell Networking OS
penalizes the interface by assigning a penalty (1024) that decays exponentially
depending on the configured half-life. After the accumulated penalty exceeds
the suppress threshold value, the interface moves to the Error-Disabled state.
This interface state is deemed as “down” by all static/dynamic Layer 2 and
Layer 3 protocols. The penalty is exponentially decayed based on the half-life
timer. After the penalty decays below the reuse threshold, the interface
enables.
Dampening settings timing example:
Lets say we have dampening 10 100 1000 60
·
half-life = 10 seconds
·
reuse-threshold = 10
·
suppress-threshold = 1000
·
max-suppress-time = 60 second
Time after flap
|
Penalty
|
Port state
|
Comment
|
0s
|
1024
|
Down
|
Penalty set to 1024
Penalty (1024) > Supress-threshold (1000) then port state down
|
10s
|
512
|
Down
|
Penalty set to 1024 / 2
Penalty (512) > Reuse-threshold (100) then
port state still down
|
20s
|
256
|
Down
|
Penalty set to 512 / 2
Penalty (256) > Reuse-threshold (100) then
port state still down
|
30s
|
128
|
Down
|
Penalty set to 256 / 2
Penalty (256) > Reuse-threshold (100) then
port state still down
|
40s
|
64
|
Up
|
Penalty set to 128 / 2
Penalty (64) < Reuse-threshold (100) then port
state is changed to UP
|
No comments:
Post a Comment