SLA ICMP-echo - how many fails before switching?

[jasonww04]

I'm using the following config for failover. My question is, how many times does the echo fail before SLA will switch the default route?

track 1 ip sla 1 reachability

ip sla 1

icmp-echo 63.123.252.1 source-interface FastEthernet4

ip sla schedule 1 life forever start-time now

ip route 0.0.0.0 0.0.0.0 173.220.150.129 track 1

ip route 0.0.0.0 0.0.0.0 64.115.219.65 200

[Tim Butters · Accepted solution]

Jason,

Once. One failed ICMP echo and the track object will be down.

You can manipulate this by using the delay up / down function within the Track object. You can specify the delay option in seconds.

I have an IP SLA frequency of 10 seconds. My Track objects have a delay down of 30. So three ping fails and the Track object will go down.

Hope this helps. Rate if you find useful.

[jasonww04]

If you change the frequency to 10, that means the SLA will happen every 10 seconds instead of 60? If you left the frequency at 60 and use the delay down 30, would the track object go down after 180 seconds (three pings)?

Would it look like this?

track 1 ip sla 1 reachability

delay down 30

ip sla 1

icmp-echo 63.123.252.1 source-interface FastEthernet4

frequency 10

ip sla schedule 1 life forever start-time now

ip route 0.0.0.0 0.0.0.0 173.220.150.129 track 1

ip route 0.0.0.0 0.0.0.0 64.115.219.65 200

[manoj.3.joshi.con]

In this case,

1. say at t1 ICMP echo failure happen.
2. Next ICMP request will trigger at t1+60 Sec.
3. track object would wait till configured delay down time (t1+30 sec in this case) and will go down. Default route switchover will take place.
4. As the delay down time is less than frequency, always route switchover will hapen after icmp failure.
5. If next ICMP(t1+ 60 Sec) is successful, default router will switch back to original. If not, it will continue to use floating route.

To avoid switchover with single flap, as a recommendation, always frequency should be set to very less value i.e. 1 to 5 Sec and delay should be set to moderately high value i.e. 10 to 15 sec.

PS: Delay value less than frequecy is bound to create switchover.

[tonyf66]

This is a specific reply to jasonww04's question at (07-27-2011 09:51 AM) about setting the frequency to 60 secs and the delay to 30 secs:

I would say it will go down on the first failed reply,  before the 180 secs, because once it goes down when set as described with (frequency 60) and (delay 30), it appears will not poll again until the "delay time" has expired; therefore, being in a down state when the "delay time" has expired would "register" as such.

[Alex Pfeil]

I think that using a frequency of 10 seconds, and then a tracking a delay down of 30 seconds is a good solution. Does the 30 seconds account for three missed pings or would it actually only account for 2. For example, should the delay down be for 31 seconds?

Does anybody use a different amount of time for tracking, or is there a best practice?