WAE replacement procedure in a WAAS-WCCP deployment at DC

[mnolden]

Dear community,

In regards to the topic below I have the following information (based on Cisco best practice documentation) if WCCP with WAAS-WAE is configured on a C6500:

WAE Add or Replacement Procedure

1.       Disable WCCP on all the WAEs. By default WCCP will wait for 180 seconds for the existing TCP sessions to close by the end hosts.

2.       Disable WCCP globally using ‘no ip wccp 61’ & ‘no ip wccp 62’ on all the routers/switches that are part of WCCP farm.

3.       Remove the failed WAE from the network.

4.       Add the new WAE and enabled WCCP version 2 on all WAEs in the WCCP farm.

5.       After all WAEs are enabled for WCCP then enable WCCP globally for WCCP service group (61 and 62 for WAAS)

My concern on this replacment procedure on C6500 is related to the fact that all WAE devices has to be removed from WCCP process before I can add a new or a replacement WAE. In a high availability environment like DC minimum one WAE has to do further on WAAS optization via WCCP redirection while the other WAE is added or replaced.

Question:

1. Can you confirm the correctiveness of the  WAE add and replacement procedure based on a WCCP deployment on C6500 switches?

Reference: WhitePaper -- Cisco WAAS WCCP on Cisco Catalyst 6500 - Best Practice Recommendations for Successful Deployments.pdf

2. If this WAE add and replacement procedure is true for C6500 switches, is it valid for ASR1000 series routers and C3750 too?

Many thanks for your comments and feedback

[Bhavin Yadav · Accepted solution]

Hello,

Yes. This process is correct for any WCCP deployment. There are various reasons behind this.

1. When you disable / enable WCCP on any one router / switch / wae, esp. with 6500, there are chances that it will corrupt the bucket assignment due to shuffling of buckets when something drops out from the farm.

2. This is true when you have multiple WAEs or routers / switches in the WCCP farm.

3. Ideally, when you take out any WAE / switch / router, boucne the WCCP.

4. When you notice that something (WAE / Switch / Router) has dropped out of WCCP farm, please boucne the WCCP.

Point to be noted: When you enable WCCP, please make sure to enable WCCP either on all routers / all switches / all waes within 10 seconds. Bucket assignment and allocation will start after initial HIA packets from WAEs to routers which have 10 seconds of timer.

The mis-alignment of buckets may cause issues like - some of the subnets not able to redirect thru WCCP, blackholing traffic for some subnet / any other odd random behaviour.

There is no need to disable WCCP on both sides. i.e. WAEs and (Routers / Switches). Disabling this on one side should be enough to restart any WCCP assignments.

Further, esp. with 6500, if there is any change in ACL or any parameters, it will affect the CPU utilization heavily and CPU will spike for few minutes while it shuffles the TCAM table.

Hence, it is always a best practice to bounce WCCP either on all routers and switches OR on WAEs to make sure WCCP aligns itself properly. This hold true for any / all switches / routers.

Hope this helps.

Regards.

PS: Please mark this as Answered, if this answers your question.