Does changing QOS policy bandwidth numbers cause an outage?

Unanswered Question
Sep 24th, 2009

I was changing the bandwidth numbers in our QOS service policies to increase the amount of bandwidth for video conferencing. Out of a dozen remote routers, this went just fine for ten of them. But on a couple of them, when I made the change, EIGRP lost adjacency across the WAN and had to reconverge - which only took a few seconds.

Not only did I increase the bandwidth amount in the video class, but I renamed the policy map, which caused me to remove it from the WAN interface an re-apply it with the new name. Here is the policy map and interface service policy:

policy-map QOS_classes_24calls_3video

class Voice

priority 2511

class Video

bandwidth 2766

class Call-Control

bandwidth 500

class class-default

fair-queue

random-detect

policy-map WAN-100MB-speed

class class-default

shape average 99920000

service-policy QOS_classes_24calls_3video

All the remote WAN routers are connected into a Metropolitan Optical Ethernet backbone, and all WAN interfaces on the WAN routers are 100MB ethernet.

Is there supposed a brief outage when a change is made to QOS policy parameters, or when service policies are removed and re-applied to interfaces? Should I expect enough outage that EIGRP could be affected?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
Loading.
Giuseppe Larosa Thu, 09/24/2009 - 08:43

Hello Jim,

by using a different policy map with a different name you did well your procedure of change.

I've noticed a long delay in telnet prompt in applying QoS to the ATM interface of a C877, I was starting to think to have missed it and then the device comes back.

Removing the service policy causes the interface to remove the SW queues, then applying another service policy causes the insertion of a new set of SW queues on the interface.

I wouldn't say you should expect 15 seconds outage that makes EIGRP neighborship to fall down.

However, some seconds of freeze are possible in my case I hadn't a routing protocol running but it was much more then 15 seconds.

It may be safe to schedule a maintanance window next time you change them even if I don't think users can perceive this.

Hope to help

Giuseppe

Actions

This Discussion