As noted in the comments, signals may not be the best way to accomplish this. I’m assuming that you do want signals nonetheless.
You can use the standard user signals: SIGUSR1
to enable maintenance and SIGUSR2
to disable it.
Use os/signal
to get notified of these signals and update program state:
// Brief example code. Real code might be structured differently
// (perhaps pack up maint and http.Server in one type MyServer).
var maint uint32 // atomic: 1 if in maintenance mode
func handleMaintSignals() {
ch := make(chan os.Signal, 1)
go func() { // FIXME: use Server.RegisterOnShutdown to terminate this
for sig := range ch {
switch sig { // FIXME: add logging
case syscall.SIGUSR1:
atomic.StoreUint32(&maint, 1)
case syscall.SIGUSR2:
atomic.StoreUint32(&maint, 0)
}
}
}()
signal.Notify(ch, syscall.SIGUSR1, syscall.SIGUSR2)
}
Have a middleware look at that state and respond accordingly:
func withMaint(next http.Handler) http.Handler {
return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if atomic.LoadUint32(&maint) == 1 {
http.Error(w, "Down for maintenance", http.StatusServiceUnavailable)
return
}
next.ServeHTTP(w, r)
})
}
You can apply this middleware on a per-route basis, or directly to the server’s root handler:
func main() {
handleMaintSignals()
srv := http.Server{
Addr: ":17990",
Handler: withMaint(http.DefaultServeMux),
}
srv.ListenAndServe()
}
You don’t need a second executable like changeMyServerStatus
. Use your operating system’s tools to send signals, such as pkill:
$ nohup myserver &
$ curl http://localhost:17990/
404 page not found
$ pkill -USR1 myserver
$ curl http://localhost:17990/
Down for maintenance
$ pkill -USR2 myserver
$ curl http://localhost:17990/
404 page not found
But manually juggling nohup
and pkill
is tedious and error-prone. Instead, use a service manager such as systemd to manage your process. Systemd lets you send arbitrary signals with systemctl kill
:
systemctl kill -s SIGUSR1 myserver