How to ignore specific query parameter with varnish cache?

/swatches/ajax/media/?product_id=17620&isAjax=true&_=1524469655019

I need to cache this request with varnish, but i need to ignore the last query parameter _=1524469655019.

I am not sure how varnish internally works, but i suppose it makes the cache uid key from the url requested. So in my case it would need it only to create the uid key from this url

/swatches/ajax/media/?product_id=17620&isAjax=true

Doing something like req.url ~ "^/swatches/(.*)$" would not work as varnish would still use the entire url for cache uid.

bitumin

2 years ago

Level 1

It seems you never got an answer. Sad story bro. But for future reference purposes... What I see often done for params like yours (and others, like the usual tracking utm params) is:

Before doing the return (hash) in the vcl_recv subroutine, strip the param(s) from the url like so (example with utm params and the _ you mentioned):

if (req.url ~ "(\?|&)(utm_[a-z]+|_)=") {
    set req.url = regsuball(req.url, "(utm_[a-z]+|_)=[-_A-z0-9+()%.]+&?", "");
    set req.url = regsub(req.url, "[?|&]+$", "");
}

(Not tested, so do not just copy-pasta...)

In this way, the stripped params are ignored for caching purposes (so, not added to the caching hash). Keep in mind that a side effect of this is that these params won't reach your app backend either, so if you need to process them for whatever reason, you might actually want to avoid the caching altogether when those params are present (which would involve to do a param presence check and a conditional return (pass), but that is another story...).

By the way, I hope the _ is not a content versioning param, because... if it is, I suspect you might not want to strip it from the caching hash 🤔

Please or to participate in this conversation.