here is the English translation: Amplify Might Be AWS's Worst Service, Bar None
Confusing documentation, a mix of old and new systems, and it made a mess of my AWS account.
To put it simply, over the past two days, I attempted to deploy a full-stack assignment on AWS services. The front end was written in React, using Vite as the framework. For such Single-Page Apps (SPAs), I personally prefer using specialized services like Netlify or Cloudflare Pages for deployment, as these services offer very robust CI/CD services, allowing for one-click deployment and automatic updates, saving a lot of hassle.
Initially, I planned to manually deploy on AWS using the S3 + CloudFront model (since it was just a one-time assignment), but later I discovered that AWS has a service very similar to Netlify called Amplify, which also offers CI/CD one-click deployment services. Amplify goes even further by including user directory services, allowing for one-click registration and login via related components.
It sounds great, but you only realize how problematic it is after using it. After some research, my initial deployment method was to upload the code to GitHub and then click the deploy button in the Amplify interface. This is also the deployment method I use most often with Netlify.
However, I later found something wrong. The key issue was that applications deployed this way using Amplify couldn't directly use Amplify's UI components to access Cognito user directory services. After much searching, I found that Amplify has an Amplify CLI initialization command to create a new CI/CD project in the Amplify service, which also deploys additional resources like Cognito.
It seemed feasible, so I did it. Then I found some issues. The initial "issues" were just on the AWS account management level: after deploying the project via Amplify CLI, my AWS account quickly filled up with a bunch of "things"—the reason "things" is in quotes is that Amplify created a lot of fragmented resources, including but not limited to CloudFormation, IAM roles, etc., even creating two Cognito identity pools for me—it's hard not to call them "junk." Moreover, most of these resources have names that are impossible for humans to remember or distinguish, and there are no explanations or grouping features to tell you what these things are for.
If it were just like this, it wouldn't seem to impact the development process, right? The biggest problem is that the local debugging and production environment apparently don't use the same configuration files, and when I was cleaning up the automatically created resources in my AWS account earlier, I somehow deleted the roles calling the Cognito user pool in the production environment, causing the production environment to be unable to access the two user pools created by Amplify, constantly throwing 400 errors.
After several rounds of "deploy-delete-redeploy-redelete," I decided to start over and look for the related documentation again. Later, I found that Amplify has a set of documentation outside of AWS's own documentation system, and this documentation recommends a deployment method: clicking the deploy button on the GUI webpage—yes, you heard it right, the same deployment method I used initially.
So, how do you deploy additional components/services like Cognito this way? Amplify's answer is configuration files. As long as you create a folder for configuration files in the root directory of your project and write the corresponding configuration files in it, the cloud will automatically create the resources you need in AWS once it reads them.
It sounds reasonable, right? Then you go to find the part about configuration files in the documentation... What's going on? Why can't I find anything in the search box in the documentation? There's not even a sample configuration file! Algolia indexing service can't be this bad, right?
Searching for "defineAuth" in the Amplify official documentation returns mostly irrelevant information.
Is my search method incorrect? I entered keywords like "site
.amplify.aws defineAuth" in the Kagi.com search engine but couldn't find any examples or explanations of configuration file items. At this point, I'm completely convinced that the Amplify documentation is garbage. Fortunately, the API documentation of the Amplify framework is quite good, at least reducing my urge to buy a ticket to the US and blow up Amazon's headquarters while guessing the configuration file items...
Also, Amplify has a UI that is completely different and more modern than other AWS services. The discrepancy is still a minor issue; the main problem is that if you create a project using the (slightly outdated) Amplify CLI, and then try to configure the back-end services like Cognito it deployed on the webpage, you'll enter an old interface. That is, once you click in, you see a slightly ugly but familiar interface, yet it feels completely disconnected from the previous Amplify interface...
So now I understand why I hadn't heard of Amplify before—it's really hard to use. Complete integration is indeed an advantage, but even being born with a silver spoon doesn't excuse Amplify's messiness, simply throwing everything together and telling users "it just works." Users look at it, wondering what on earth all these things are, and then you hand them a manual that looks fancy but has zero information. Users, flipping through this tome with no useful information, can only throw this pile of stuff into the historical junk heap behind them in frustration.
Some rights reserved
Except where otherwise noted, content on this page is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International license.
ENJOY
READ THIS
Short story: "The Cry of Accelerated Demise" in a place accelerating towards demise.