john-osullivan
commented
5 years ago Closing this issue, as we're no longer using invalidations in our update strategy. Instead the post-build action will trigger an update to the Cloudfront distro's default root object, instantly triggering a cache miss. This is both cheaper and faster than using invalidations. Progress on the new strategy is on this other issue: https://github.com/Eximchain/abi-clerk-lambda/issues/38
Currently, the CloudFront invalidation is created during the update operation. If that processes before the build, and someone accesses it before the build, it could re-cache the old content.
We should have the invalidation trigger after the pipeline build is complete. We could trigger either on the pipeline completion (If that's possible) or on modifications to the Dapp S3 bucket.