This is yet another follow up to post to several previous posts about using nginx as a reverse proxy with caching. It is actually a direct addition to my post from a week or so ago which outlined how to actually using nginx’s proxy caching feature which can be read here — /2010/02/07/nginx-proxy_cache-and-explained-benchmarked/.
Even more changes?
Yes, even more changes, these are basic changes that are there to improve the caching capabilities and also implement load balancing.
Cache changes
The first set of changes are in the main nginx configuration file
/etc/nginx/nginx.conf
These changes basically just change the proxy_cache key
proxy_cache_path /var/www/nginx_cache levels=1:2 keys_zone=cache:8m max_size=1000m inactive=600m;
proxy_temp_path /tmp/nginx;
proxy_cache_key "$scheme://$host$request_uri";
I’ve decided to put the temporary caches file in to an nginx specific directory, just to separate them from other cache files. I’ve also modified the proxy_cache_key to add the following variables:
- $scheme - This will be the protocol; http or https
- $host - Host name, this will be set as syslog.tv for me
- $request_uri - The full request uri, this is simple
Why add these variables? Quite simple really, it means I can have multiple sites running with proxy caching enabled and have them set a meaningful key so they don’t clash.
Load balancing changes
The next set of changes will be specific to my site specific configuration file, the first of which is an addition on an upstream definition.
/etc/nginx/sites-available/syslog.tv
upstream apachesyslogtv {
server apache.syslog.tv weight=1 fail_timeout=60s;
}
So what does this mean? This is really actually quite simple, we define an upstream set called apachesyslogtv, which contains, in this example, a single server definition with a weight of 1 and a fail timeout of 60 seconds. You would actually be able to add multiple server definitions to this with different weights and fail timeouts. This is used for load balancing.
upstream apachesyslogtv {
server apache1.syslog.tv weight=1 fail_timeout=60s;
server apache2.syslog.tv weight=1 fail_timeout=60s;
}
Server definition changes
There are quite a lot of changes that are made to the actual server definition, I will go through these step-by-step to explain what has changed and why it’s changed.
location / {
root /path/to/site;
proxy_redirect off;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
if ($http_cookie ~* "comment_author_|wordpress_(?!test_cookie)|wp-postpass_") {
set $do_not_cache 1;
}
proxy_cache_key "$scheme://$host$request_uri $do_not_cache";
proxy_cache cache;
proxy_cache_valid 200 302 60m;
proxy_cache_valid 404 1m;
proxy_pass https://apachesyslogtv;
}
Ok, so that’s the first location, definition. Several things have changed with this.
The first change is an addition of a cookie check, in my case I’m running a WordPress site so I check for various WordPress cookies, if they exist I set the variable $do_not_cache to 1.
Next is a change to the proxy_cache_key to incorporate the $do_not_cache variable, this tells nginx not to cache the current page.
And finally is a change to proxy_pass, this now points to the name of my upstream definition, which nginx will then use to decide which server to use.
location ~* .(jpg|png|gif|jpeg|css|mp3|wav|swf|mov|doc|pdf|xls|ppt|docx|pptx|xlsx)$ {
proxy_cache_valid 200 120m;
expires 604800;
proxy_pass https://apachesyslogtv;
proxy_cache cache;
}
The second and final location definition, which will match a file extension in the URL, if a match is found it will set the cache validity to 2 hours, expire to 7 days, pass back to our upstream definition and cache the result. You may notice that out of all of these static file extensions .js is missing, this is because a lot of my site is generated by Javascript/Ajax and this will not work with caching.
Why the changes?
With these changes I am able to properly store with a cache key which allows me to cache all of my nginx sites, I’ve added the ability to balance load across multiple servers and I’ve increased the power of the caching to only cache pages if you’re not logged in and to always cache static files for a long time. Meaning that the cache shouldn’t need to be regenerated very often.
All in all these changes do not increase the power of the server at all, but with a load balanced environment this would obviously increase the performance dramatically.
The full config
upstream apachesyslogtv {
server apache.syslog.tv weight=1 fail_timeout=60s;
}
server {
listen 174.143.241.61:80;
server_name syslog.tv;
access_log /var/log/nginx/syslog.tv.access.log;
gzip_vary on;
gzip_static on;
location / {
root /path/to/site;
proxy_redirect off;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
if ($http_cookie ~* "comment_author_|wordpress_(?!test_cookie)|wp-postpass_") {
set $do_not_cache 1;
}
proxy_cache_key "$scheme://$host$request_uri $do_not_cache";
proxy_cache cache;
proxy_cache_valid 200 302 60m;
proxy_cache_valid 404 1m;
proxy_pass https://apachesyslogtv;
}
location ~* .(jpg|png|gif|jpeg|css|mp3|wav|swf|mov|doc|pdf|xls|ppt|docx|pptx|xlsx)$ {
proxy_cache_valid 200 120m;
expires 604800;
proxy_pass https://apachesyslogtv;
proxy_cache cache;
}
}
