Primary Fail Event POST

Hi! Cant wrap around how to configure haProxy.
Main idea is, i have main server and backup server, if main is down i need to post few request to startup the backup server, this has to be done just once, until i manually return primary to working state and reload haproxy.
I find this approach configs, witch doesn’t suite, because it returns the lua response instead of backup backend response.

global
    log /dev/log local0
    log /dev/log local1 notice
    daemon
    lua-load t.lua  

defaults
    log     global
    mode    http
    option  httplog
    timeout connect 5000
    timeout client  50000
    timeout server  50000

frontend front_9001
    bind *:9001
    default_backend back_9001
    http-request lua.log_primary_status

backend back_9001
    mode http
    option httpchk GET /api/v2/BaseInfo
    http-check expect status 200
    option persist
    option allbackups
    server primary 10.0.0.35:9001 check fall 3 rise 2 inter 5s shutdown-sessions
    server backup 10.0.0.36:9001 backup check

and in lua for test for example

function log(message)
    local file = io.open("/var/log/haproxy_status.log", "a")
    file:write(os.date("%Y-%m-%d %H:%M:%S") .. " - " .. message .. "\n")
    file:close()
end

-- This function will be called with each request to log the primary server status
core.register_action("log_primary_status", { "http-req", "tcp-req" }, function(txn)
    local backend = core.backends["back_9001"]
    local primary_server = backend.servers["primary"]

    -- Check if the primary server is down
    if primary_server and primary_server:get_state() == "DOWN" then
        log("Primary server is down - Health check failed.")
    else
        log("Primary server is up.")
    end
end)

-- Log when the Lua script is loaded
log("Lua script loaded. Version: " .. os.time())

So is there a way to execute lua method in here

server primary 10.0.0.35:9001 check fall 3 rise 2 inter 5s shutdown-sessions

to startup the backup itself, log alert and so on with lua, and then remember that we did that, and even if the primary is back alive, return responces from backup untill manual restart?

Hello,
To follow the discussion.
I have the same problem for one of my projects.