Problem with cache and urls with query strings

[luisherranz]

Hi mchung,

I'm doing some tests and it seems like it's caching urls which it shouldn't.

If you load http://www.mydomain.com/?a=1 and then you turn off the database, you still can load http://www.mydomain.com/?a=1 http://www.mydomain.com/?a=2 http://www.mydomain.com/?a=3... which obviously you shouldn't, because urls with queries aren't supposed to be cached.

Can you confirm it?

Best, Luis.

[mchung]

Hi Luis-

Can you explain what's supposed to happen..

• What do you mean by turn off the database? Wouldn't this just takedown the site?
• Why wouldn't URLs with a query string be cached? Is there a reference doc you can point me to about how that's supposed to work?

[luisherranz]

Sure,

1. To know if a page is cached or not, I visit it and then turn off the database. If it loads properly, it's cached. If I get an "Error establishing connection with the database" it's not.
2. Look at these lines from the original wordpress.conf.erb: # POST requests and urls with a query string should always go to PHP if ($request_method = POST) { set $cache_uri 'no cache'; } if ($query_string != "") { set $cache_uri 'no cache'; }

The second if means that nginx shouldn't cache urls with query strings. This is the default Wordpress behavior when you use cache plugins like WP SuperCache or W3 TotalCache.

It is useful when your Wordpress uses queries to do php things, like an affiliates system which uses a tag to identify each affiliate: ?aff_id=X.

If you don't need it you can always comment those lines, but now they are not working.

[luisherranz]

Ok, there's a simpler method to check this:

Load www.mydomain.com and look at the source. In the header you should see something like: <!-- generated in 1.398 seconds 59036 bytes batcached for 300 seconds -->

If you reload it you should see something like this: <!-- generated 28 seconds ago generated in 1.398 seconds served from batcache in 0.020 seconds expires in 272 seconds -->

Now, if you load something with a query string, like www.mydomain.com/?something=1, it shouldn't generate a cache, but it does. You still get: <!-- generated in 1.279 seconds 59078 bytes batcached for 300 seconds --> And if you reload: <!-- generated 23 seconds ago generated in 1.279 seconds served from batcache in 0.017 seconds expires in 277 seconds -->

For some reason, it is caching urls with query strings.

By default, batcache shouldn't do that. This is from the batcache documentation:

Exemptions # Note that URLs with query strings are automatically exempt from Batcache. This can be undesirable in many cases as popular pages linked to with query strings can significantly reduce the effectiveness of our caching setup and can affect the overall performance of your site.

And with the current wordpress.conf.erb configuration it shouldn't either: # POST requests and urls with a query string should always go to PHP if ($request_method = POST) { set $cache_uri 'no cache'; } if ($query_string != "") { set $cache_uri 'no cache'; }

I'm doing tests, trying to change things but I don't seem to find the problem.

[luisherranz]

I have removed these lines: # Don't use the cache for logged in users or recent commenters if ($http_cookie ~* "comment_author|wordpress_[a-f0-9]+|wp\-postpass|wordpress_logged_in") { set $cache_uri 'no cache'; } but even with that it doesn't use cache for logged in users, so this is not working at all.

I think I am completely lost here.

[luisherranz]

Ok, so the lines referring cache on the wordpress.conf.erb doesn't have any effect.

I've finally made it work editing the $args nginx variable and advanced-cache.php. Maybe it's not the best way to deal with this, but it is working fine so far.

I will try to explain it here:

• Javascript-only queries: These are the ones you want to be ignored when you serve cache, but you want to be passed so the js scripts can do their work.
  • For example: www.mydomain.com?javascript_tag=123456 should serve the cache of www.mydomain.com
  • These are typically tags of AdWords (gclid), Google Analytics (utm_campaign, utm_source...)
  • You need to remove them from the nginx $args variable, like this: if ($args ~ ((.*)(\?|&)javascript_tag=\w+$|(.*)&?javascript_tag=\w+&?)(.*)) { set $args $2$4$5; }
• Non-cache queries: These are the ones you don't want to be cached at all. If they exist, cache won't be served.
  • For example: www.mydomain.com?no-cache=123456 should be load always dynamically.
  • These are typically queries used by php to modify the html or database in some way.
  • You need to add them in advanced-cache.php, after this $batcache = new batcache($batcache); if ($_GET['no-cache'] != '') { return; }
  • I suppose that if you don't want to cache anything with query strings by default you could use something like: if ($_SERVER['QUERY_STRING'] != '') { return; } but I haven't tested it myself.
• Cache queries: These are the ones you want to be cached.
  • For example: www.mydomain.com?cache-me=123456 should serve www.mydomain.com?cache-me=123456
  • You don't need to do anything, this is the default behaviour.

[mchung]

Luis, can you visit the page from a new browser session, say incognito mode (never logged in, no cookies, etc). In my early testing, I also came across this weird behavior where basically the caching only applied to visitors. All the caching would be skipped when the user was logged in to the admin site.

[luisherranz]

mchung, it's working fine here, it never caches when I'm logged in.

Can you look in the code? If the page is cached you should see something like this: <!-- generated 4 seconds ago generated in 1.358 seconds served from batcache in 0.012 seconds expires in 296 seconds -->

[mchung]

Luis- What did you want me to look into?

[luisherranz]

I think I misread your last post, sorry. Everything is working fine here.