Reset session on my Virtual Server when a specific pool member with high priority come UP

What service type are you providing (HTTP, other) and what monitor are you using?

Hi cjunior,

i use for both pool member a standard icmp monitor (default for all node) and for mypool a "TCP-Half open" port monitor dedicated to the daemon port to monitor. The service is not http,but is an application that establish a connection via TCP socket.

Thanks

Within pool, if you set "Reset" as an option for "Action on Service Down", does it help ?

Hi Cryosleep,

the easiest way would be a setup using...

1. Enable Priority Group Activation prefering ServerA
2. Attaching a OneConnect Profile to your Virtual Server to enable LTM to rebalance existing TCP-Connections on a per HTTP-Request basis (requires HTTP Profile)
3. Remove any Persistence Profile from your Virtual Server.

By doing so, any existing connection to ServerA will failover to ServerB in the case ServerA is unhealthy. And any existing connection to ServerB will failback to to ServerA in the case ServerA is become healthy again.

No iRule or scripting required... ;-)

Cheers, Kai

Hi Cryosleep,

you may try one of the iRules below, to rebalance active sessions for ServerB if ServerA comes back online...

iRule 1: Using a per TCP-Packet

CLIENT_DATA

when RULE_INIT {
    set static::IP_ServerA "10.10.10.1%1"
    set static::IP_ServerB "10.10.10.2%1"
}
when SERVER_CONNECTED {
    if { [LB::server addr] equals $static::IP_ServerB } then {
        clientside {
            TCP::collect
        }
    }   
}
when CLIENT_DATA {
    if { [active_members -list [LB::server pool]] contains $static::IP_ServerA } then {
        LB::detach
        LB::reselect
        TCP::release
        return
    }
    TCP::release
    TCP::collect
}

iRule 2: Using a periodic TCL script on every active connections to ServerB to lookup the availability of ServerA and then failover.

when RULE_INIT {
    set static::IP_ServerA "217.110.108.152%1"
    set static::IP_ServerB "217.110.108.152%1"
    set static::reject_interval 5000 ; msec
}
when SERVER_CONNECTED {
    if { [LB::server addr] equals $static::IP_ServerB } then {
        after $static::reject_interval { 
            if { [active_members -list [LB::server pool]] contains $static::IP_ServerA } then {
                after cancel -current
                LB::detach
                LB::reselect
            }
        }
    }
}

Note: iRule 1 should be considered as very agressive for your CPU, since the status of ServerA is checked on each single TCP-Packet send from the client to ServerB (allows very fast failbacks). iRule 2 uses a once every X msec interval to check the status of ServerA for server side connection send to ServerB (less CPU hungry approach).

Note2: Both iRule are designed to keep the client side TCP-Connection active by transparently reselecting the server side side connection from ServerB to ServerA. If this is a problem for your application, then report back and I'll rewrite the iRule, to reject the client side connection.

Cheers, Kai