How to set up a failover node/pool?

This should do the trick...

when HTTP_REQUEST {
  if { [active_members "def_pool_name"] == 0 } {
    pool "fallback_pool_name"
  }
}

Just replace the "def_pool_name" and "fallback_pool_name" with your default and fall back pool names.

-Joe

Hi Joe,

 

 

The script you given can switch the traffic to standby pool if the active pool fail. Then, how to start a application server and its service that was not running at a standby node? For our case, we also need to remount shared storage to the stadby node at the same time. Are there any way that can restart application server at a standby node before switching traffic? Thanks.

 

 

John

The only thing I can think of would be to build a special command handler on your servers and then when you detect you need the services restarted, modify the payload to the backend server to trigger this operation. Then on response to the BIG-IP, you could reset the original request and issue a HTTP::retry to send the original request to that secondary server.

something like this comes to mind:

when HTTP_REQUEST {
  set ORIG_REQ [HTTP::request]
  set $needs_reset (decision logic goes here)
  if { $needs_reset } {
     Change URI to special reset command on server
    HTTP::uri "/reset_services"
  }
}
when HTTP_RESPONSE {
  if { $needs_reset } {
     Revert back to original request and retry
    LB::reselect
    HTTP::retry $ORIG_REQ
  }
}

You may have to issue a HTTP::collect in the HTTP_REQUEST if you have a large request payload and then process the logic in the HTTP_REQUEST_DATA event instead.

Hopefully this gives you some food for thought.

-Joe

To elaborate... what about going back to the default pool - would this logic work?

 

 

when HTTP_REQUEST {

 

if { [active_members "def_pool_name"] == 0 } {

 

pool "fallback_pool_name"

 

}

 

elseif { [active_members "def_pool_name"] == 1 } {

 

pool "def_pool_name"

 

}

 

}