Nginx caching strategy based on request

Question

Lets say my backend takes a horrific 2 seconds to respond (out of our control say), and I never EVER want a visitor to experience this delay, we are not influenced by cookies or user session.

What I want to do in nginx is something like this:

server { location / { # server requests from cache unless for some reason its not in cache proxy_pass http://backend/ proxy_cache mycache proxy_cache_is_valid 2m } location / where request IP = 127.0.0.1 { proxy_pass http://backend/ do_not_use_cache but_store response in cache } } 

This way I can have a simple curl task from localhost run every 30 seconds that keeps the cache fresh/hot with the several pages I need, I never want the visitor to be the one that warms up the cache. I've read the docs but cant see how to make this happen.

Answer 1

Try this configuration.

http { # http section directives proxy_cache_path /path/to/cache levels=1:2 keys_zone=mycache; geo $bypass_by_ip { default 0; 127.0.0.1 1; } server { # other server directives location / { proxy_cache mycache; proxy_cache_valid any 2m; proxy_cache_use_stale timeout updating; proxy_cache_bypass $bypass_by_ip; proxy_pass ...; } } } 

proxy_cache_bypass makes a direct request, bypassing the cache. This is controlled by its argument(s) -- if any of them is not an empty string or 0. I am using geo to provide such a value (0 by default and 1 when remote IP is 127.0.0.1)

NOTE you need Nginx version 1.0.4 at least for this configuration to work. Earlier versions had a bug in proxy_cache_bypass / proxy_cache_nocache logic.