BIGIP LTM Loadbalancing for Cache Farm with Client IP Spoofing feature.

Dear everyone.

Please help me with this topology.

Network topology attached as below.

Routing:

Cache farm using Bluecoat Cache Devices.

Default route of caches is BigIP VLAN 2’s ip address (233.x.x.129)

Default route of BigIP is Vlan 3 on Multilayer Switch (MS) (233.x.x.230)

Vlan 1 of MS connect to Subscribers network and Vlan 2 connect direct to Internet router.

Configurations:

BIG-IP LTM:

I created two Pools:

Go_in_pool: has 3 members are three Cache devices (233.x.x.130-132)

Go_out_pool: has only one member: VLAN3 MS (233.x.x.230)

And two Wildcard Virtual Server associate with 2 pools:

Go_in_server:

- Destinations: Network 0.0.0.0/0.0.0.0

- Service Port: 0

- Type: Performance L4

- Protocol: All Protocol

- Enable on: VLAN 1 (external)

Go_out_server:

- Destinations: Network 0.0.0.0/0.0.0.0

- Services port: 0

- Type: Performance L4

- Enable on: VLAN 2 (internal)

Cache Device:

- Caching for HTTP Traffic

- Using Client IP Spoofing feature: This feature allow Cache devices sent HTTP Request to real server with IP address of client.

Multilayer Switch: Redirect traffic:

Rule 1:

- Flow: Vlan 1, ingress

- Source IP address: Subscribers IP address

- Destination port: 80

- Action: Redirect to 233.x.x.229 (BIG-IP SelfIP Vlan external)

Rule 2:

- Flow: Vlan 2, Ingress

- Destination IP address: Subscribers IP address

- Source Port: 80

- Action: Redirect to 233.x.x.229 (BIG-IP SelfIP Vlan external)

And now I have serious problem.

- When Client sent HTTP requset (www.example.com), MS redirect this request to F5. F5 sent to Cache1.

- Cache 1 check itself, if it has content for this request, it return for Client via F5 – This’s OK.

- If Cache 1 does not has content, it act as client and sent request to real server (www.example.com) to get data.

- But when using Client IP Spoofing Feature, Cache sent request with source address of Client.

- When Real server reply, it sent packet HTTP response with destination IP is IP address of Client via Internet Router.

Internet Router sent this packet to MS. MS redirect this packet to F5.

At this point. F5 see this package as new request and sent to Cache farm

- IF F5 send this package to Cache1, Cache 1 has content and response to Client correctly.

- IF F5 send this package to Cache 2 or Cache 3, Cache 1 wait to time out for it’s connection. And after time out, it send an error response to Client.

That’s why our network too slow and has a lot of error sessions.

Please give me an advice to fix this case.

Do you have persistence turned on?

CB

Dear CB.

I tried enable persistence based on source address, dest address, … but it seem to bad, client access internet more and more slowly.

When I disable persistence, Client go faster but there’re alot of error session.

Destination persistence is what is usually recommended for Proxy caches. I don’t think that would be the root cause, but it could be pointing to another issues. Curiously is your network connection on the LTM and blue coat set to auto or is it hard coded?

CB

Dear CB.

Both network connection on LTM and Blue Coat Caches set to auto.

I think problem is way F5 process traffic. Example, when Cache 1 sent request to www server (@ internet) with source IP address is Client IP address, and www server return with destination IP is Client IP address. At this point, Content Switch redirect traffic to F5, but I think F5 sent this traffic flow to another Cache. That’s why internet connection to slow, cause Cache 1 still waiting response of www server.

Rereading your problem again I was thinking why would you need the client IP spoofing turned on. In other words why would you want to send the IP address of yoour client to the internet server?

CB

Our network is an ISP. So we must enable Client IP Spoofing feature. If not, some download site like rapidshare, megaupload will limit quota for Cache IP, and then, client clould not download from these sites.

Could you please check if the auto last hop feature is enabled. This feature is found in system->configuration->Local Traffic->General.

This feature would make sure that the return traffic is sent to the same cache which initiated the connection.

an news for this issue?

Just waiting back for a response post from duongnt.

CB

Dear all.

Finally, I got support from F5 Technical and we have to use iRule.

I post this iRule here for reference.

when HTTP_REQUEST {

set High_Score -9999999999

set Node_Picked “”

set Pool_Picked pool_[expr {crc32 [HTTP::uri]} % 10]

foreach Cur_Node [active_members -list $Pool_Picked] {

if { [crc32 $Cur_Node[HTTP::uri]] > $High_Score } {

set High_Score [crc32 $Cur_Node[HTTP::uri]]

set Node_Picked $Cur_Node

}

}

pool $Pool_Picked member [lindex $Node_Picked 0] [lindex

$Node_Picked 1]

}

Note that this iRule requires version 9.4.2+. Change to

your pool name.

And for versions prior to 9.4.2

Change "myPool" to your pool name. 
  Replace IPADDR1, IPPADDR2, IPADDRN with the node IPs contained in "myPool" 
  Replace  with the port number of the nodes contained within "myPool" 
  
 when HTTP_REQUEST {  
 set GOOD_NODES {} 
 set High_Score -9999999999 
 set Node_Picked “” 
 set Pool_Picked pool_[expr {crc32 [HTTP::uri]} % 10]  
 set NS {IPADDR1 IPADDR2 IPADDRN } 
 set portnum  
    This looks through myPool nodes to determine the one's that are active 
    and passes it onto the array called GOOD_NODES 
   foreach N $NS { 
         if { [LB::status pool  $N $portnum] eq "up" } { 
            lappend GOOD_NODES $N 
         }       
    } 
  
 foreach Cur_Node $GOOD_NODES { 
     if { [crc32 $Cur_Node[HTTP::uri]] > $High_Score } { 
     set High_Score [crc32 $Cur_Node[HTTP::uri]] 
     set Node_Picked $Cur_Node 
     } 
  } 
  pool $Pool_Picked member [lindex $Node_Picked 0] [lindex $Node_Picked 1] 
 }

CB