How to switching URI roundrobin based on data-group list

Create 2 pools. POOL-1 contains 120, 121 and 122 with round robin method. POOL-2 contains 119.

when HTTP_REQUEST {
if { [HTTP::host] starts_with "apisky" } {
pool POOL-1
} else {
pool POOL-2
}
}

Hi swjo,

to overwrite the HTTP::host value based on the currently selected pool member (aka. just for pool member X.X.X.119), you have to use the

HTTP_REQUEST_SEND

event (triggered after pool member selection), filter on the selected pool member IP and finaly change the HTTP::host value to a random value based on a mod(rand())^4 calculation followed by an array lookup to retrive the HTTP::host names...

when RULE_INIT {
    array set static::apisky {
        "0" "api31sky.whypaymore.co.kr"
        "1" "api32sky.whypaymore.co.kr"
        "2" "api33sky.whypaymore.co.kr"
        "3" "api34sky.whypaymore.co.kr"
    }

}
when HTTP_REQUEST_SEND {
    if { [LB::server addr] equals "X.X.X.119" } then {
        clientside {
            HTTP::host $static::apisky([expr { int( rand() * 100 ) % 4 }])
        }
    }
}

Note: The selection of the individual HTTP::host values in the example above is not a true "Round-Robin-Selection". Instead its more or less just a "Random-Selection". To implement a true "Round-Robin-Selection" you have to store a shared counter value into LTMs session-

[table]

so that each TMM core can access and also increase/rollover this information as needed. But in 9 out of 10 times a real "Round-Robin-Selection" isn't required and should therefor be avoided to save CPU cycles...

Cheers, Kai

Hi Expert, Please correct me. If you have list of real client IP address, Then create IP list in data group, match with iRule & redirect to node 119.

 

And rest connection should move to pool_A, where all the nodes are available.

 

Does it makes sense?

 

Hi swjo,

To get a "true" Round-Robin mechanism for pool member selection and HOST-Header overwrites you have to implement some shared

[table]

-based RR counters, so that the individual TMM-cores (each of them will process a subset of the client connections) can become teamed. The shared RR counters will then be used to perform a handcrafted RR pool member selection followed by the RR based injection of the HOST-Header values.

Note: You have to perform a handcrafted pool member selection, since LTM's build-in "Round Robin" load balancing algorythm does not share the RR counters across multiple TMM cores. It will just Round-Robin the request accordingly with that specific TMM instance...

when RULE_INIT {
    array set static::apisky_hostheaders {
        "0" "api31sky.whypaymore.co.kr"
        "1" "api32sky.whypaymore.co.kr"
        "2" "api33sky.whypaymore.co.kr"
        "3" "api34sky.whypaymore.co.kr"
    }
    set static::apisky_hostheaders_count [array size static::apisky_hostheaders]
}
when HTTP_REQUEST {

     Fetching a list of the currently available pool members

    if { [set pool_active_members [active_members -list [LB::server pool]]] ne "" } then {

         Increasing the global RR counter for load balancing decission

        set pool_rr_counter [table incr "pool_rr_[LB::server pool]"]

         Reset the global RR counter for load balancing decission as needed.

        if { $pool_rr_counter > 1000000 } then {
             The RR counter has exceded the maximum allowed value. Reset it back to 1
            set pool_rr_counter [table set "pool_rr_[LB::server pool]" 1] 
        }

         Performing a Load-Balancing decision based on the global RR counter value for load balancing decission and the list of currently available pool members.
        set pool_selected_member [lindex $pool_active_members [expr { ( $pool_rr_counter - 1 ) % [llength $pool_active_members]}]]

         Using the RR selected pool member as the current node

        pool [LB::server pool] member [lindex $pool_selected_member 0] [lindex $pool_selected_member 1]

         Increasing the global and per-pool_member RR counter for HOST-Header selection

        set member_rr_counter [table incr "member_rr_$pool_selected_member"]

         Reset the global and per-pool_member RR counter for HOST-Header selection as needed.

        if { $member_rr_counter >= 1000000 } then {
             The RR counter has exceded the maximum allowed value. Reset it back to 1
            set member_rr_counter [table set "member_rr_$pool_selected_member" 1] 
        }

         Resolve the HOST-Header based on the values of $static::apisky_hostheaders() by selecting a modulus of the current RR counter value
        HTTP::host $static::apisky_hostheaders([expr { ( $member_rr_counter - 1 ) % $static::apisky_hostheaders_count }])

         Log the current node selection and choosen HOST-Header value
        log local0.debug "Current Node: $pool_selected_member | New HOST-Header: [HTTP::host]"  

    } else {
         The pool is unavailable...
        HTTP::respond 503 content "The pool is currently unavailable..." "Content-Type" "text/html"
    }
}

Log File:

Aug 28 13:47:13 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:13 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:13 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:14 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:14 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:14 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:14 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:15 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:15 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:15 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:15 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:15 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:16 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:16 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:16 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:16 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:16 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:16 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:17 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:17 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:17 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api31sky.whypaymore.co.kr
Aug 28 13:47:17 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:17 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:17 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:18 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api32sky.whypaymore.co.kr
Aug 28 13:47:18 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:18 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:18 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:18 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api33sky.whypaymore.co.kr
Aug 28 13:47:19 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 1.1.1.1%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:19 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 2.2.2.2%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:19 f5-02 debug tmm[25655]: Rule /Common/my_rule_rule : Current Node: 3.3.3.3%1 80 | New HOST-Header: api34sky.whypaymore.co.kr
Aug 28 13:47:19 f5-02 debug tmm1[25655]: Rule /Common/my_rule_rule : Current Node: 4.4.4.4%1 80 | New HOST-Header: api34sky.whypaymore.co.kr

Note: You can clearly see that pool members as well as the HOST-Header values are switched in a "true" Round-Robin manner, even in the case that multiple TMM core will procress this connection (tmm and tmm1 in my case).

Cheers, Kai