对于使用cloudfront在S3上托管的静态站点，如何使index.html的缓存无效？

Question

So I have hosted my angular app on s3 with a cloudfront dist. I do file revision-ing (using grunt filerev) to make sure that I never get stale content. But, how should I version the index.html file. Its required because all other files are referenced inside index.html.

I have configured my bucket to be used as a static site. So it just picks up the index.html when I reference the bucket in url.

Cloudfront says that you should set min TTL to 0, so it would always hit origin to serve the content. But, I don't need this since I am doing file revisioning of all my files (except index.html). I can take advantage of cdn caching for these files.

They also say that in order to invalidate a single object, set the max-age headers to 0. I tried adding following to my index.html

<meta http-equiv="Cache-Control" content="public, must-revalidate, proxy-revalidate, max-age=0"/>

But this does not reflect once you upload on s3. Do I need to explicitly set headers on s3 using s3cmd or dashboard? And do I need to do this every time when index.html changes and I upload it?

I am aware that I could invalidate a single file using cmd but its a repeating process and It would be great if it can take care of itself just by deploying on s3.

Answer 1

Although the accepted answer is correct if you are using s3cmd, I was using the AWS CLI, so what I did was the following 2 commands:

First, to actually deploy the code:

aws s3 sync ./ s3://bucket-name-here/ --delete

Then, to create an invalidation on CloudFront:

aws cloudfront create-invalidation --distribution-id <distribution-id> --paths /index.html

Answer 2

Answering my own question. I deploy my site to S3 using s3cmd tool and there is an option you could provide to invalidate CloudFront cache of all the files changed (diff between your dist folder and S3 bucket). This invalidates cache of all the files changed including index file. It usually takes around 15-20 mins to reflect the new changes on production.

Here is the command

s3cmd sync --acl-public --reduced-redundancy --delete-removed --cf-invalidate [your-distribution-folder]/* s3://[your-s3-bucket]

^{Note: On macOS, you can install this tool via: brew install s3cmd .}

Hope this helps.

Answer 3

You can automate a process using Lambda. It allows you to create a function that will perform certain actions (Object invalidation in your case) in response to certain events (new file in S3).

More information here: https://aws.amazon.com/documentation/lambda/

Answer 4

When you sync local directory with s3, you can do this:

aws s3 sync ./dist/ s3://your-bucket --delete

aws s3 cp \
   s3://your-bucket s3://your-bucket \
   --exclude 'index.html' --exclude 'robots.txt' \ 
   --cache-control 'max-age=604800' \
   --metadata-directive REPLACE --acl public-read \
   --recursive

The first command is just a normal sync, the second command enable S3 to return cache-control for all the files except index.html and robots.txt .

Then your SPA can be fully cached (except index.html ).

Answer 5

If you use s3cmd sync and utilize the --cf-invalidate option, you may have to also specify --cf-invalidate-default-index depending on your setup.

From the man page:

When using Custom Origin and S3 static website, invalidate the default index file.

This will ensure to also invalidate your index document, most likely index.html, which will otherwise be skipped regardless if updated or not through the sync.