(assets): force re-generation of assets via the CLI

[nija-at]

We hit a bug in NodeJS version that generated corrupted assets.

The problem is that just changing to a non-buggy version of NodeJS is insufficient. This is because the assets are cached in the cdk.out/ directory and cached in the staging S3 bucket.
To resolve this issue, both has to be manually deleted.

This is a request to provide an option via the CLI where all asset caches are ignored and asset packaging/staging is re-run.

---

This is a 🚀 Feature Request

[github-actions]

This issue has not received any attention in 1 year. If you want to keep this issue open, please leave a comment below and auto-close will be canceled.

[diranged]

I'd like to re-open this... we are running into this bug relatively frequently with our integration tests. I suspect we're running into #23290 somehow with the integration test runner, but it's causing a ton of pain for us. We would really like to be able to disable the cache for builds.

[jk2l]

I would like to re-open this too. I encountered this bug in prod environment #30967

One of our developer done a prod release via --exclusive flag, this caused the asset bundled without dependency, our lambda was completely broken, because of the CDK ignore asset upload if it already exist. Both of CodePipeline or local CDK cli won't fix the issue.

At the end I need to manually go into S3 bucket hunt down each one of those lambda zip file in the staging asset bucket, and delete them manually. This is not really an ideal situation, lucky we only affected very few lambda during this deployment, but imagine if i have a deployment that trigger lot of lambda this will be a nightmare to fix (e.g. i recently deployed 1000 lambda update accross 70 accounts, because Python3.7 retired with EOL)