Thanks for your options

> 3) Use a heartbeat solution - http://linux-ha.org/wiki/Main_Page

I have API between the Nginx and the second (control) server. I use RPC interface. I need just send a RPC request (with status infrmation) to the control server. I think that heartbeat is too robust.

> 4) Use cron to start a script which does the test, sleeps for 30s, then does
> the test again, then dies (to be restarted by cron
> at the next minute)

This option seems to be the best but is any possibility the nginx to do it in its own?