Ember项目部署至AWS S3/CloudFront未同步变更问题求助

Hey there, let's figure out why your Ember app's updates aren't showing up in S3 even though your deployment says it succeeded. I've run into similar issues before, so here are the most likely culprits and how to fix them:

1. Your deployment script isn't syncing all files correctly #

This is the most common issue. If you're using the AWS CLI's aws s3 sync command, double-check two critical details:

• Missing --delete flag: By default, sync only uploads new/modified files but doesn't delete old assets from S3. Ember generates new hashed asset filenames on each build, so old assets will linger, and if your script isn't pushing the new ones (or skipping index.html), you won't see changes. Add --delete to your sync command to clean up old files and ensure all new files are uploaded.
• Incorrect source directory: Make sure your script is pointing to the latest dist folder generated by ember build --prod. It's easy to accidentally deploy an old build if you forgot to re-run the build command before deploying.

If you're using ember-cli-deploy-s3 or a similar plugin, check your config/deploy.js to confirm:

• The bucket name matches your target S3 bucket
• filePattern: '**/*' is set to ensure all files are included
• The environment is set to production (so it uses the correct build output)

2. S3 is skipping file uploads due to metadata checks #

Sometimes S3's sync logic might skip files if it thinks they haven't changed, even if they have. This often affects non-hashed files like index.html (which Ember doesn't hash by default). To fix this:

• Add the --exact-timestamps flag to your aws s3 sync command. This forces S3 to compare exact file modification times instead of just size/ETag, ensuring newer local files overwrite old ones in S3.
• Check if your deployment tool is setting custom cache headers that interfere. For example, if index.html has a long Cache-Control header set, some tools might skip updating it unless explicitly told to.

3. You forgot to re-run the production build #

It sounds obvious, but double-check: did you run ember build --environment production (or ember build --prod for short) before deploying? Local development builds (ember build without flags) use different asset paths and don't optimize files, so deploying that old dist folder won't reflect your latest changes. You mentioned seeing changes in local assets, but confirm the dist folder you're deploying from is the one generated by the production build.

4. Permissions or silent deployment failures #

Occasionally, deployment tools report success even if they hit permission issues. Verify that the IAM user/role you're using for deployment has these S3 permissions:

• s3:PutObject (to upload files)
• s3:DeleteObject (to remove old files if using --delete)
• s3:ListBucket (to compare local and S3 files)

Run your deployment command with verbose mode (e.g., aws s3 sync --verbose) to check logs for skipped or failed file uploads.

Quick Troubleshooting Step #

Manually compare your local dist folder with the S3 bucket via the AWS Console:

• Look for new hashed asset files (e.g., app-abc123.js) — if they're missing in S3, your script isn't uploading them.
• Check the modification time of index.html in S3 vs. your local dist — if the S3 version is older, the file wasn't overwritten.

Once you fix the S3 sync issue, don't forget to invalidate your CloudFront cache (if you haven't set up automatic invalidations) — even if S3 is updated, CloudFront might still serve cached old content.

内容的提问来源于stack exchange，提问作者Nick Shcherba