cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1992
Views
0
Helpful
3
Replies

2 WAE WCCP l2 only 1 gets the traffic

Hi,

I have 1 WAN Router and 2 WAVE devices configured in WCCP. The configuration works fine except that only the first WAVE that sees the router and established the WCCP receives the traffic. What I mean is that both WAVEs see the router and vice versa. When I establish the WCCP connection, the first WAVE to establish it becomes LEAD WAE and the other one does not get packets. If I disconnect the lead WAE or change its WCCP config and put it back, WCCP switches over to the other WAE and the other one is now exclusevly receiving the traffic. No load balancing is acheived.

First here's my setup:

1 WAN Router Cisco ISR G2 2911 IOS 15.2(1)T

2 Cisco WAVE-274 WAAS version 4.3.3 configured identically for WCCP.

Router IP: 10.x.y.1/22

WAVE IPs: 10.x.y.9 and 10.x.y.7 /22 and default gateway is the router 10.x.y.1

Users are on the same network 10.x.y.0/22 (is this a problem? i read in some WAAS config guide that the WAE cannot be in the same network as users)

Second here's the relevant config:

Router:

ip cef

ip wccp 61

ip wccp 62

interface GigabitEthernet0/0

description *** LAN Connection ***

ip wccp 61 redirect in

ip addr 10.x.y.1 255.255.252.0

!

interface GigabitEthernet0/1

description *** WAN Connection ***

ip wccp 62 redirect in

ip addr WAN_IP...

!

WAAS:

primary-interface GigabitEthernet 1/0

interface GigabitEthernet 1/0

ip address 10.x.y.9 255.255.252.0 (and .7 for the second WAVE)

!

interface InlineGroup 1/1

shutdown

!

wccp router-list 1 10.x.y.1

wccp tcp-promiscuous router-list-num 1 l2-redirect l2-return

wccp version 2

When I do the following on the router:

show ip wccp 61 detail

or show ip wccp 62 detail

I see:

WCCP Client information:

        WCCP Client ID:          10.x.y.7

        Protocol Version:        2.0

        State:                   Usable

        Redirection:             L2

        Packet Return:           L2

        Assignment:              HASH

        Initial Hash Info:       00000000000000000000000000000000

                                 FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF

        Assigned Hash Info:      FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF

                                 00000000000000000000000000000000

        Hash Allotment:          128 (50.00%)

        Packets s/w Redirected:  103912

        Connect Time:            03:34:05

        GRE Bypassed Packets

          Process:               0

          CEF:                   0

          Errors:                0

        WCCP Client ID:          10.x.y.9

        Protocol Version:        2.0

        State:                   Usable

        Redirection:             L2

        Packet Return:           L2

        Assignment:              HASH

        Initial Hash Info:       FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF

                                 00000000000000000000000000000000

        Assigned Hash Info:      00000000000000000000000000000000

                                 FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF

        Hash Allotment:          128 (50.00%)

        Packets s/w Redirected:  0

        Connect Time:            01:46:24

        GRE Bypassed Packets

          Process:               0

          CEF:                   0

          Errors:                0

On the WAAS, the WCCP Assignment Settings for Load Balancing is the default: Hash. (Hash on Source IP (Service 61):)

the Egress Method is IP forwarding

I have several connections from different source IP addresses and somehow they all end up hashed on the same WAE:

ConnID        Source IP:Port          Dest IP:Port            PeerID          Accel     RR

   360          10.x.y.3:49463   10.q.w.36:52732      xx:xx:xx:xx:xx:xx TMDL  16.1%

   373          10.x.y.4:55005   10.q.w.36:52732      xx:xx:xx:xx:xx:xx TMDL  24.8%

I checked in several places and read the best practices; the router platform support... and it seems that the config is OK

http://www.cisco.com/en/US/prod/collateral/contnetw/ps5680/ps6870/white_paper_c11-608042.html

Any ideas?

Thanks,

Patrick


1 Accepted Solution

Accepted Solutions

This questions got discussed in the 'ask the expert' session:

https://supportforums.cisco.com/message/3482826

View solution in original post

3 Replies 3

Although it is recommended to use HASH for 2900 series routers, I just switched to MASK method for load balancing on both WAVE devices. This is supported according to Cisco doc. It seems that connections are now being accelerated by both WAVEs.

The behaviour is a bit weird though, connections are first being sent to one WAE then they show up as passthrough on it for a quick second and after that they get treated by the second wave!

I also see this behaviour when looking at the counters on the router, the counters went up to 274 packets on one router and are no longer changing although new connections are being treated, while the other router has a lot more packets:

ROUTER#show ip wccp 61 detail

WCCP Client information:

        WCCP Client ID:          10.x.y.7

        Protocol Version:        2.0

        State:                   Usable

        Redirection:             L2

        Packet Return:           L2

        Packets Redirected:      274

        Connect Time:            01:49:58

        Assignment:              MASK

        Mask  SrcAddr    DstAddr    SrcPort DstPort

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

        0000: 0x00000F00 0x00000000 0x0000  0x0000

        Value SrcAddr    DstAddr    SrcPort DstPort CE-IP

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

        0008: 0x00000800 0x00000000 0x0000  0x0000

        0009: 0x00000900 0x00000000 0x0000  0x0000

        0010: 0x00000A00 0x00000000 0x0000  0x0000 

        0011: 0x00000B00 0x00000000 0x0000  0x0000 

        0012: 0x00000C00 0x00000000 0x0000  0x0000 

        0013: 0x00000D00 0x00000000 0x0000  0x0000 

        0014: 0x00000E00 0x00000000 0x0000  0x0000

        0015: 0x00000F00 0x00000000 0x0000  0x0000

        WCCP Client ID:          10.x.y.9

        Protocol Version:        2.0

        State:                   Usable

        Redirection:             L2

        Packet Return:           L2

        Packets Redirected:      100788

        Connect Time:            01:49:56

        Assignment:              MASK

        Mask  SrcAddr    DstAddr    SrcPort DstPort

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

        0000: 0x00000F00 0x00000000 0x0000  0x0000

        Value SrcAddr    DstAddr    SrcPort DstPort CE-IP

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

        0000: 0x00000000 0x00000000 0x0000  0x0000 

        0001: 0x00000100 0x00000000 0x0000  0x0000 

        0002: 0x00000200 0x00000000 0x0000  0x0000 

        0003: 0x00000300 0x00000000 0x0000  0x0000 

        0004: 0x00000400 0x00000000 0x0000  0x0000

        0005: 0x00000500 0x00000000 0x0000  0x0000 

        0006: 0x00000600 0x00000000 0x0000  0x0000 

        0007: 0x00000700 0x00000000 0x0000  0x0000 

Any ideas?

Maybe I should've just clustered the WAVEs inline...

This questions got discussed in the 'ask the expert' session:

https://supportforums.cisco.com/message/3482826

Although the router, both WAEs and the users at the site were in the same L2 domain; WCCP + L2 redirect was not the method to use. I moved from L2 redirect to GRE and that fixed the issue. Thanks to Peter's suggestion.

We later changed the setup by trunking the router's LAN interface and moving the WAEs to a "management" vlan that is seperate from the users LAN network.