I have JS and CSS files being served through Amazon Cloudfront. Up until now I was using my own server running Varnish, and when I updated any JS or CSS file or I had to do to use that new version was to increase a counter on a URL param on the request for that file.
So the URL of the file being included would go from …/file.css?r=1 to …/file.css?r=2 and Varnish would think of that as a new file, request it and cache it and serve it from cache from that point on.
With Cloudfront I’m realizing it doesn’t work like that, it still serves the old cached file (and I keep getting “hit from Cloudfront” on the HTTP response).
Any idea if I can change the behavior of Cloudfront for this?
Anything else I can do that doesn’t imply actually changing the name of the file everytime I update the file? I also don’t want to use cache invalidations through Cloudfront’s API. I think that’s too much work for what I’m trying to do here, and there’s a limited amount of (free) requests per month I can make.
Looks like you were about a month too early… in mid-May of 2012, Amazon announced that you could start using query strings to identify objects in your cloud front cache.
Here’s a link to the announcement on their blog: http://aws.typepad.com/aws/2012/05/amazon-cloudfront-support-for-dynamic-content.html
Unfortunately, I don’t see any links in my quick googling that describes how exactly you can do this but I’m sure it’s buried in their documentation or control panel somewhere.
Hope this gets you going in the right direction.