Cache Busting when more than 1 web server into load balancer

[dbpolito]

Because of this:

https://github.com/cartalyst/themes/blob/3.0/src/Assets/AssetManager.php#L475

Cache names are different across the web servers, and as we the default way of load balancer, it splits all requests between server, in my use case i'm getting the html from one server, but it's trying to get the assets from another, which doesn't exists because the cache busting are different.

I don't want to have a CDN for now, so would be good have some cache busting configuration...

[mmic-bjohnson]

In the OP's case, and if memory serves, the $cacheKey was different every time because:

1.) The paths to the files were different on each server.

2.) The last-modified timestamps were different on each server.

Both issues can be resolved with some changes to the deployment process (e.g., ensure that paths are the same on all servers, by whatever means appropriate, and synchronize timestamps with rsync).

That said, it would be nice to be able to control the caching behavior with more granularity.

@dbpolito What types of options would be helpful to you?

[dbpolito]

Well, elixir doesn't have this problem... I don't know the strategy for generating hashes, if it uses md5 of the file or something, but elixir hashes are the same across the servers...

1. The path is different because envoyer symlink the build folder to the current folder, so /path/to/projeto -> /build/TIMESTAMP , this can be easily fixed by stop using the absolute path and use the app root path, starting from the /public

2. The build is done concurrently, but it's basically impossible to the files has the same timestamp

Take a look at the elixir version strategy for generating hashes, it just works...