How does Nginx handle the request of the upstream server when it is marked as `down`?

[Maxim Dounin]

Hello!

On Mon, Apr 13, 2015 at 04:16:56AM -0400, HUMing wrote:

> For a simple Nginx configuration like this:
> For a simple Nginx configuration like this:
> 
> upstream myservers {
>     server 127.0.0.1:3000;
>     server 127.0.0.1:3001;
> }
> server {
>     listen       80;
>     location / {
>         proxy_pass http://myservers;
>     }
> I have two questions related zero downtime of the application:
> 
> If I change the configuration to mark the first server server 127.0.0.1:3000
> as down, I assume that no new request will go to that server, but what about
> the current request that is handled by the upstream server? Does Nginx can
> still return valid response to end user for that request?
> 
> If I remove the first server server 127.0.0.1:3000 and reload the
> configuration, what about the current request that is handled by this
> upstream server?

Both removal of the server and marking it down are equivalent as 
long as you are using round-robin balancing (there is a difference 
when using ip_hash and hash balancers, as "down" implies less 
remapping when temporary disabling a server).

In both cases the server won't be used by new worker processes to 
handle new requests, and old worker processes will gracefully 
terminate upon completion of previously started requests.

See here for details:

http://nginx.org/en/docs/control.html#reconfiguration

-- 
Maxim Dounin
http://nginx.org/