IP SLA Issues

Unanswered Question
Jan 2nd, 2009
User Badges:

Hi all,


I'm having an issues with a IP SLA I created which is controlling a static route. It works but is a little unreliable.


ip sla 1

icmp-echo 87.85.165.76 source-interface Vlan2 (this is a static host I use)

timeout 20000

threshold 2

frequency 30

ip sla schedule 1 life forever start-time now


Periodically when i run "sh track" it shows last change x minutes ago rather than days. The overall status appears to be "up" however.


Track 100

Response Time Reporter 1 reachability

Reachability is Up

15 changes, last change 00:03:23

Latest operation return code: Over threshold

Latest RTT (millisecs) 39

Tracked by:

STATIC-IP-ROUTING 0


Is there a way I can make this less sentative - I would ideally like it to only declare "down" after 30 secs with timeout of 20000 to rule out a flapping link?

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
The_guroo_2 Fri, 01/02/2009 - 08:24
User Badges:

mate you have put the frequency 30 ......its in seconds so after every 30 second it will query.....change that and you will be fine


hope it helps

rasoftware Fri, 01/02/2009 - 08:31
User Badges:

Posted by: rasoftware - 244 Fulwood road

Jan 2, 2009, 8:05am PST



Hi all,


I'm having an issues with a IP SLA I created which is controlling a static route. It works but is a little unreliable.

I originally had it set to 15 but it seemed worse.


Do I understand this correctly -


Target- icmp-echo 87.85.165.76 source-interface Vlan2

Timeout - timeout 20000


Number of failures - threshold 2


Every x seconds - frequency 30


So every 30 secs, request ping within 20000ms, if two fail (threshold) - declare it down?


I always see to have this line in sh track


Latest operation return code: Over threshold





MARTIN STREULE Tue, 01/06/2009 - 04:16
User Badges:
  • Silver, 250 points or more

Use the delay command in the "track" part of your config:


delay (tracking)

To specify a period of time to delay communicating state changes of a tracked object, use the delay command in tracking configuration mode. To disable the delay period, use the no form of this command.


delay {up seconds [down seconds] | [up seconds] down seconds}


no delay {up seconds [down seconds] | [up seconds] down seconds}



More info:

http://www.cisco.com/en/US/docs/ios/iproute/configuration/guide/irp_prb_mult_track_ps6441_TSD_Products_Configuration_Guide_Chapter.html


Hope that helps.


Cheers,

Martin

rasoftware Tue, 01/06/2009 - 04:21
User Badges:

Hi Martin,


Thanks I worked this one out - it's been up for 3 days now without reporting a change using the delay up 10 sec down 30 secs.


Stability is key for my config and I only want failover if it been down for a minute or more.


It is much more stable now.


Thanks for the reply.

Pravin Phadte Tue, 01/06/2009 - 06:36
User Badges:
  • Silver, 250 points or more

Hi,


Your Question:

Is there a way I can make this less sentative - I would ideally like it to only declare "down" after 30 secs with timeout of 20000 to rule out a flapping link?


Declare down after 30 seconds:

Command "frequency 30"


Rule a flapping link ?

Commadn " timeout 20000" (not sure that what is defined"


-----


Remove the static route if the link is down more than 30 seconds and not flapping.


The flapping can occur in minutes and you link may be up for 5 min and then again down or may be down in 1 min 2wice.


------------

This works great for me 5 min.

Paramites needed:

num-packets 3

frequency 300


remove

timeout 20000

threshold 2


When the route goes off. This will wait for 5 min down to replace the route.


So if your link flaps in this 5 min it will shift to backup unless it gest 3 packests in 5 min probe.


Once its down. It changes it route.


Now best part it will stay on the other route unless gain it recives prove of 5 min stable 3 packtes from the icmp-echo ip.


so if the link flaps on icp-echo ip provided it will not relace the route. Once the link is stable it will replace the route.


Downtime not more than 5 sec,


But for the first swicth over it depends on the frequency.


do decide about the frequesy you need to set.


Hope this helps


Regards,


pravin


rasoftware Tue, 01/06/2009 - 07:34
User Badges:

Thanks Pravin,


I think i've cracked it with the 20000ms, freq 30sec, threshold 2.


Adding delay down 30 sec up 10 has prevented the status changing.


I believe if it missing two pings of a very high latency in 60 sec - it then starts count down - so a max 90 sec before failover. This seems to prevent "flapping" of the route due to a few missed pings when the link is really fine.




The sh track now corresponds to the router uptime, before I was getting a status change every 5 mins or so when the link was fine.


Since I have a VPN working over it - this was causing the tunnel to drop etc which is not good for the terminal users behind it.



I will investigate the "num packets" line as this might provide further more accuracy. I suspect my one just sends a single ping each 30 secs rather than 3 which should provide less false positives

Pravin Phadte Tue, 01/06/2009 - 23:08
User Badges:
  • Silver, 250 points or more

Thats good to know that the problem is solved.



georges.merhej Tue, 02/03/2009 - 03:06
User Badges:

Hello Guys,


Do you have any idea on how to clear the "show track" statistics?


MOE#show track

Track 1

Response Time Reporter 1 reachability

Reachability is Up

57 changes, last change 10:14:49

Latest operation return code: OK

Latest RTT (millisecs) 40



Thanks,


Georges

Pravin Phadte Tue, 02/03/2009 - 04:18
User Badges:
  • Silver, 250 points or more

hi,


copy the output for show run sla monitor xxx


config t

no sla monitor xxx


check track it will be 0


then paste back the configs for sla monitor 123

Actions

This Discussion