Confused on IP SLA

[harry]

Hi all,

I have a router configured a fail-over between  the primary Link which is VDSL (Interface Eth0) and the backup link 4G (Interface Cellular0). I use IP SLA to track the primary as below. I know that the default timeout value of ip sla is 5000 ms which means when the router send out a icmp-echo, it will wait for a response up to 5 seconds before declaring the primary link down (Correct me if I'm wrong at this point). 

I did a ping test to 110.145.65.1 (My Internet Gateway) from the router itself and recognized that the maximum round trip time of a ping probe was never longer than 2 seconds. However, IP SLA kept flapping between the 2 links which resulted in an unstable internet connection for LAN users. It seems for me the timeout value doesn't work properly. so what is the reason for that? 

What are the correct values of Timeout, Frequency or Delay down that I should set in IP SLA to only trigger a fail-over if a ping probe fails for 10 seconds?

ip sla 1

icmp-echo 110.145.65.1 source-interface Ethernet 0

frequency 5

ip sla schedule 1 life forever start-time now

track 10 ip sla 1 reachability

ip route 0.0.0.0 0.0.0.0 Ethernet0 110.145.65.1 track 10

ip route 0.0.0.0 0.0.0.0 Cellular0 5

Thanks,

[Richard Bradfield · Accepted solution]

Hi these are the typical settings I use and don't get any problems

ip sla 1
icmp-echo 110.145.65.1 source-interface Ethernet 0
frequency 30
ip sla schedule 1 life forever start-time now

track 1 ip sla 1 reachability
 delay down 40 up 10

Ping every 30 seconds, track delay 40 so get 2 ping failures before say it is down.

HTH

[harry]

Hi Richard,

Thanks for your response.

Does "delay down" overpower "timeout" in IP SLA? As I know the default timeout is 5000 ms.

and how did you work out the number of ping failures which is 2 in this case?

For me, a delay of 40 seconds is quite long. I prefer 10 seconds. Also can please you tell what "delay up 10" does?

Thanks

[Richard Bradfield · Accepted solution]

the delay should be longer than the timeout so you know the ping has
failed.You can adjust the timeout


*(config-ip-sla-echo)#timeout ? <0-604800000> Timeout in milliseconds*
So say you ping every 5 seconds and adjust the timeout to 3 seconds
then lets say a ping fails and we set the tracking delay to 9 seconds, so
that ensures that 2 consecutive pings have failed before you say it is
down, and it will come up after the delay in the up time after a
successful ping.

[harry]

Thanks Richard,

It is clear for now. I'll test it on my client's router next week. 

[harry]

Hi Richard,

As Cisco's recommendation
"Apart from the IP SLAs UDP jitter operation. For all other IP SLAs operations, the following configuration guideline is recommended:
(frequencyseconds ) > (timeoutmilliseconds ) > (thresholdmilliseconds )"
So I wonder that is there any potential problem if I set timeout value equal to frequency value?
For example:
ip sla 1
icmp-echo 110.145.65.1 source-interface Ethernet 0
frequency 5
timeout 5000
ip sla schedule 1 life forever start-time now

[Ludovic Kuty]

Cisco gives the following recommendation about the timeout value in its IP SLA Command Reference: "We recommend that the value of the milliseconds argument be based on the sum of both the maximum round-trip time (RTT) value for the packets and the processing time of the IP SLAs operation." It makes sense since it is the time after which we can reasonably conclude that the destination is unreachable.

It is nicely combined with the track delay which let us specify after how much "unreachable time" using IP SLA we should conclude that the track is down and thus that we have to react to correct the situation. We don't want to react too fast otherwise we will have a lot of Down/Up track state changes.