how to modify my wccp load-balance

[chia hao chang]

hi all

I've two WAE in my core, 612 (10.99.100.95) and 674 (10.99.100.96)

and 3750 stack is my core switch

I configured wccp tcp-promiscuous router-list 1 l2-redirect l2-return mask-asign in both WAE

and wccp load balance is work fine

but in my environment ip subnet, like 10.10.10.0

10.10.10.1 ~ 10.10.10.128 will load balance in WAE-612

and 10.10.10.128 ~ 10.10.10.255 will load balance in WAE-674

and I want to reverse it

I try to change WAE ip address

but it still haven't change

anyone could help me ?

thanks

[Zach Seils]

There is no reliable way to accomplish what you are asking.  The bucket allocation is done on a best effort basis, so even if we could swap the buckets, there is no guarantee they would stay that way over time.

The best approach here is to modify the WCCP mask to get a better distribution of load across the WAAS devices.  You mention in your post that these two devices are part of your "core".  Are there any clients located at the same site with these devices, or are all of the connections being initiated from outside this location?

Zach

[chia hao chang]

HI Zach

I disable WCCP and waitting for all connection gone in WAE-674

and enable WAE-674 WCCP to join WCCP group again

it's strange

WAE-674 starts taking over the connections from WAE-612

like WAE-612 have 800 connection and WAE-674 have 400 connections before

when WAE-674 rejoin to WCCP

WAE-674 have 800 connections and WAE-612 have 400 connections now

I don't know why this principle but this is user's desire

I'm confuse@@

[Zach Seils]

This isn't necessarily a problem.  WCCP uses a hash table of 256 buckets to distribute traffic across the devices in a service group.  When you have two devices in the service group, each one gets 50% of the buckets (i.e. WAE1 gets buckets 0-127 and WAE2 gets buckets 128-255).  When a device leaves the service group, the buckets it was allocated are reassigned to the remaining device in the service group.  When a device is added to the service group, it will get 50% of the buckets, but it's not guaranteed that it will get the same 50% it previously had.

So in your case, when the 674 rejoined the service group, it's possible that it was allocated the buckets that the 612 was previously handling the 800 connections on.  This is normal behavior and not anything to be alarmed about.

Regards,

Zach