I think I was wrong about AWS Amplify
2024 ж. 4 Мам.
53 906 Рет қаралды
buckle up, this one is a long rant
00:00 aws amplify complaints
13:03 sst overview
29:32 random rant about vercel
29:58 random rant about OSS
📘 T3 Stack Tutorial: 1017897100294.gumroad.com/l/j...
🤖 SaaS I'm Building: www.icongeneratorai.com/
💬 Discord: / discord
🔔 Newsletter: newsletter.webdevcody.com/
📁 GitHub: github.com/webdevcody
📺 Twitch: / webdevcody
🤖 Website: webdevcody.com
🐦 Twitter: / webdevcody
I wanted to say that Amplify isn't awful. I made this video after spending hours debugging, and I was a bit pissed off. I'm assuming the developers working at AWS and Amplify are trying their best to make a good product; sometimes internal processes at a corporation get in the way of shipping features, and the bugs and features WE as users find important might not be prioritized inside the company due to various reasons. Amplify is still a good solution to deploy a next application, and my ONLY complaint so far was this cookie cache issue. If I wasn't using google analytics, I would have never thought twice into investigating all of these cloud front misses. This issue made me realize that abstracting away your infrastructure comes with a cost. If you are interested in having more control over your AWS infrastructure, maybe check out SST, CDK, and Terraform. IaC is another really fun aspect of cloud computing I recommend people at least try at some point once they progress in their skills.
I think you're justified in feeling pissed off. Oftentimes, if something is broken or not working, we as developers feel like we did something wrong, which leads us to investigate every point of failure. When you realize that it was the system you were leveraging, paying for, there's no work around and no idea when it's fixed, it's really frustrating.
Spent 3 hours yesterday trying to understand why I couldn’t get even the empty nextjs app working on Amplify. Turns out I had to downgrade to an older version. I’ve literally opened up a lot of tabs with github issues and searched for mentions of versions, and just averaged it out. Seems like each version hits a quirk for a specific edge case for people (besides the fact that the latest just throw a 500). I was a bit disappointed on the move to wall it into a managed service, I was expecting an SST-like experience. So thanks for the showcase, and keep up the good content! You got my sub
As a newbie who tries to build full-stack applications, I like your honest, realistic approach and process of thinking with respect to price. I'm learning many things from your channel!
sure thing, glad to hear my videos have helped
You have no idea how valuable your channel is for up and coming junior devs like me. Would love to see a cost-breakdown of AWS costs using SST.
This was a brilliant walkthrough of your troubleshooting. Awesome stuff!
Really looking forward to future videos on how you are using SST
Love this! Didn't even know I wanted to see cloud architecture debugging-style tutorials! Keep it up!
Hey Cody! Thanks for bringing this up. Your concerns have been heard by our Amplify team!
Thanks Erik. And let me know if there was something I just am not understanding because I’d hate to have a video complaining about a service when I personally just did something wrong.
It takes a team to suck this hard?
Was there a follow up about this from the Amplify team?
@@teamvashmmo3218 it’s on their road map
and any updates? 😅@@WebDevCody
I love you are going through real issues. i learned a lot from your video. anyhow i hope they'll get it fixed soon.
Amazing work! Learned new debugging techniques just from this video. Thank you 🙏
This was so detailed and in-depth. Loved hearing your insights on AWS Amplify
Great video. Nice to see real-life examples and how you think through solving problems. I don't use any AWS services but was thinking of exploring it. Biggest fear is setting things up wrong and being over charged.😅
I check my billing dashboard daily and setup billing alerts so they email me when I reach a threshold. For example, I had it set at $1 and it warned me about me hitting that threshold which made me start looking into all of this 🤣
@@WebDevCody wow I'm glad I saw this definitely going to use that tactics 😂 Plus I love your penny saving mindset I don't think many KZheadrs share things with that mindset so it's nice to see there's someone out there who shows people all this
@@amasoharnold4762 haha yeah I'm always trying to host my stuff for as little money as possible. If you don't keep track of these expenses, they can quickly grow out of control and you'll be stuck with an expensive system you can't easily change; it'll be coupled to services that slowly increase their prices every year.
@@WebDevCodysmart and frugal
This video seems to be very informative and valuable to many developers. Thank you for taking the time to go through all of this and sharing it. It's impossible not to subscribe after watching.
Thanks man glad to hear that
Really helpful video. Please keep us informed
appreciate the amount of hours you put into debugging this and sharing it with us 🙌
Great analysis on the all the "easy to use" services. Often times simplicity of use of a services comes with the price of less customizable product. Always great to know what are the limitations of a product. It is exactly what makes a good engineer in my opinion, knowing the limitations and deciding on the better product depending on the need. Always looking forward to your new videos. Cheers
Wow. Very cool debugging skills! Thank you for sharing!
thanks dude. Potentialy saved me some money. It hasn't been solved till now.
I loved your in depth on this video. I think I understand AWS and NextJs a little more, thanks to you
SST is one of the coolest tech I've come across recently. Totally democratizing you to host on AWS without being tied to Vercel. I'm seeing numerous videos popup recently about people moving away from Vercel for various reasons, one among them is bandwidth charges after you consume the free tier.
I've only played with SST for like an hour, but so far it seems nice. I've had issues in the past dealing with CloudFormation (and I know SST uses CDK which generates cloudformation templates), so that makes a bit concerned. We use terraform at work for all of our AWS resources, but I was hoping to have a cheaper "all batteries included" solution, and amplify was original meeting my needs until I hit this snag. This snag made me remember why having complete control over your infrastructure is useful; I can fix the issue in 10 minutes by changing the cache key in cloudfront using SST, vs Amplify I have to wait for their team to add in a feature that allows me to change it.
Its been a while since I modified the cloudfront dists at my work, but I believe you can set up edge functions to modify inbound and outbound requests? You could use those to strip cookies out at the cf end rather than in amplify
Great analysis! You came up with a really insightful hypothesis (cookies are included in the cache key) but I’m surprised your next step wouldn’t have been to test that hypothesis by temporarily turning off GA to see if that made a difference.
I did, just not on this video
I came across SST a few months ago and it's been great so far
Man, I'm not a junior dev, and this video simple saved my life, thank you so much.
i just also learned about SST last week! can you please make more videos related to SST?appreaciate it.
yeah I plan to
Great insight
Really enjoy these debugging videos. IMO this is what real day to day looks like in the industry and these are insights I would find useful since one day I would probably run into this and have to go down the rabbit hole myself. Having chatGPT generate that query was a real cool trick ill add to the toolbox.
chatgpt saved me hours trying to figure out how to write insight logs 🤣
This is interesting solely because you are doing something no one on KZhead has done at least that I haven't found which is DEBUGGING AWS. Bro how has this video not blown up
I’m sure others, like myself, would love to see a follow-up using SST with the T3 stack. As far as subsequent deploys- was each subsequent deploy with SST any faster? I’m wondering if the entire stack is destroyed and recreated each time you deploy.
Nah on future deploys it’s much faster. It just takes time to setup the initial distribution
Damn, who would have thought that a presumably boring video with debugging and charts would be introduced and presented so interestingly? Nice one, mate! n1!
Hey Cody, nice video as always. what do you think about Firebase? In your case for your icon generator can you deploy it(the same app) in Firebase? what are you thoughts about it?
I’ve never used it
Good job babe!!
wife located
I have been working in the cloud for many years now, and even though Amplify offers everything pre-packaged, I still prefer to do everything in Terraform! I understand that some developers prefer to have something up and running, however, most of the other options seem to cause issues for me later on. For example, we would like to implement Kinesis Data Analytics to transform the data being passed through as then you will have to work out, how can we use the resources that have been already created by the developer, which may or may not exist at a certain time as they may have destroyed it locally to push up a new change. I do like the idea behind Amplify itself being a quick setup however Google, AWS and Azure can be quite costly if you do not have good control and monitoring, especially if you are using other services such as API gateway (this has to be managed carefully) - I think if there was a way you could adjust the Cloud Formation template with additional custom resources or changes you would like to the deployed resources, then I would revisit that as it would give a little more control. I have not heard of SST until today, although it does look a little clearer, it looks like you only get the same level of control, for example, one thing you would want to add to your website is a Web Application Firewall in front of CloudFront to help with inspection and management of traffic, all this should really be possible already as it's all using Cloud Formations under the hood! I cannot see in either Rate Limiting API requests or introducing EventBridge which is used so often in serverless architecture. Someday this may all be introduced and I cannot wait as the quicker we can get developers set up the better!
SST is built on top of CDK, which uses cloudformation under the hood. So you can easily configure anything in SST simply by dropping down to the CDK layer to provision your resources. I'd give it a shot if you're already familiar with cloud formation.
@@WebDevCody Can you add this additional configuration in SST (to automatically adjust the cloud formation template) or is it that you have to build a change set later on?
12:30 maybe a decent solution would be to move the static files to a seperate subdomain (which cookies are never set for) so cache will always be hit for the static assets
Thank you for your SST and Amplify videos. My next project could potentially get more traffic, and I am afraid of the costs of vercel. So I am looking into Vercel, SST, Amplify and you mentioned Digital Ocean App. Now, one year later, what is your current recommendation for hosting? Does Vercel really get that expensive (database will be Neon), is Amplify going to be cheaper even without caching. Or go bare-bones to AWs with SST?
honestly I'm just hosting on a VPS at this point. A single VPS can host your next.js application and handle enough traffic until your application grows. I've been hosting my apps on railway and using cloudflare cdn + tunnels to protect them.
Can you do a similar video for vercel
In regards to the GA cookie, that value is your clientID that is generated to track you as an individual user, so that value shouldn't be changing every time you refresh the page, so that's weird to begin with. The clientID should remain unchanged across sessions to help GA tie your sessions together as the same user. I could be wrong though.
Probably has some blocker that prevent the cookie from staying put
well I'm seeing 2 cookies, one I think is the id that tracks you as a user, the other one seems to have a timestamp in it which probably tracks your session. Each time you load up a website, that's a new session which google analytics tries to track to understand how you use the site for the first time each session.
@@WebDevCody ngl i am running almost everything dockerized with ECS EC2 auto scaling Spot instances, which at first did sound complicated but Everytime I see more and more struggles with the serverless architecture due to all those hyped experimental features of next or meta framework of the week
What is the current vscode interface and icon pack name?.Thank you
I'm learning AWS and I came across amplify, which is like syntactic sugar to use buckets, lambda, etc. Do you recommend I continue using amplify or use pure s3 and lambda? thanks!
I'd use SST or connect directly to services over amplify personally
@@WebDevCody thanks!
after using aws for a while i feel like their only worth service is their emails,but i didnt hit the outside of free tier mark yet so probably i'll get surprise there aswell.
Can’t you also use docker and k9s to deploy to eks?
I think a request with different cookie should not be cached since its a different request, sometimes you need to change the response based on cookie value, but I do think that there should be a option to specify what cookies should be ignored. I wonder if vercel deals with this problem.
Idk, I guess it depends. Imagine having 10 images on your landing page, but since your users have different auth cookies, cloudfront never caches those images. Those images SHOULD be cached, and there isn't a reason NOT to cache them unless you built out a super dynamic system which shows different image based on the user profile. Same with the javascript chunks that are loaded from next static assets. Those do not change per user, so caching based on a cookie makes no sense to me. If I use cloudfront directly, I can white list cookies. So I could have easily added the analytics cookies to an ignore list and be done. Amplify doesn't expose that configuration for us.
The cache max age you saw is meant for your browser, right?
What would u recommend now? The best way to deploy next js apps
I like using sst if you want to stay in the aws ecosystem.
6:30 The cache control header is for instructing the client (your webbrowser), not Cloudfront itself. So the misses on the static content make perfect sense. Let me know if anything is unclear. :)
I could be wrong, but usually how cloudfront works is it checks the cache control headers returned from your origin server, and it’ll use those if they are provided depending on your cloudfront configuration of course.
@@WebDevCody Aah, I didn't know that. That makes sense. Thanks!
Indeed, the cache control header is also to be used by caching proxies between the user and the origin.
Havent really watched the full video neither ever used aws amplify. But the cookie behaviour is kind of expected and to my understanding is how it should work. In your cloudfront distribution you can enable different behaviours depending on the path the user is requesting. meaning that you could setup a behaviour for, lets say /assets/ and disable cookies forwarding, that way caching should always be a hit for images or whatever you got under your static content. For the markup, if the user is logged in, is gonna be difficult to hit cache, unless you generate the general markup that does not change if authenticated, then clientside perform whatever changes you need. Otherwise you cant really cache much, kind of depends on your use case.
Right, but I have no control in amplify to change the cache strategy of cloudfront, that’s my complaint. Right now all my images, static js files, svgs, etc and are not being cached between users because of the cookies
@@WebDevCody That seems quite strange to me that you cant manage cloudfront. Dont you have access to the cloudfront distribution via AWS console? seems quite pointless to have a cloudfront distribution if you dont have a way to tailor to your needs :(
@@MrPolistation maybe if you use the amplify cli, but I was using the amplify dashboard and they abstract all that stuff away. I’ll need to spend more time investigating everything, but with SST I managed to get it all setup real easy
FYI using GA and not having a cookie banner to disable them is illegal in the EU under the GDPR.
There's something I'm unsure about: how does client-side caching play into this? Are the cloudfront headers still updated even though the document is cached on the client?
I’m not sure what you mean
Ah sorry, maybe it's stupid, but let me try to explain it: you hadn't disabled caching on the client, meaning that in theory your files should have been served from the browser cache after the initial load. When this happens I'm unsure if headers will be refreshed?
@@DragonRaider5 in the video I think I was just explaining and I didn’t disable cache, but later in the video you’ll see I have the disable cache checkbox checked and I do a test.
@@WebDevCody ah I assumed it was enabled all along after I first saw it, thanks!
Holy Shit I thought I was crazy, I hosted a next app on amplify and I was wondering why the latency was so bad. I could have sworn that cloudfront was not involved at all, but I had no idea how to test that theory out.
Do you use cookies? I mean, I’m getting 30ms when I deploy a single next app with no cookies and cloudfront caches everything, so it’s super fast from what I’ve seen other than this small issue I have
Hello!! You weren't wrong about amplify, you just made SST stuff manually. Just for the record, you could've deactivated caching by cookies on your cloudfront policy, that is a very easy step to follow on the CloudFront distribution. If you have translations, I recommend caching by a header or cookie that has the location. And If your app has sessions, cache by the session cookie if you have one.
You can’t see your cloudfront distribution when you use amplify hosting
@@WebDevCody really? I thought Amplify just deployed and managed the services for you but you still have access to them
@@diegoquiroz1059 amplify the cli does, amplify hosting does not. It’s a very confusing service with overloaded names. Go try to deploy from your git repo in amplify and you’ll see you have no access to anything
does anyone know the name of the vscode interface with such an icon, not just me
bearded icons
It's NOT a you issue. RESPECT.
Also amplify does not handle SSR right? its just static output you host ?
It does ssr
DID THEY FIX THE COOKIE CACHE PROBLEM?
Idk I stopped caring
why not aws lambda?
What we learned today: Do not RECOMMEND technologies, say that you "love" a thing, before REALLY USING! I know, everyone make mistakes...
Why not try azure?
to lazy to learn another cloud provider. aws has paid my bills so I rather just keep using it
thats why its better to use rust aws sdk and use rust for making lamda , its very very cheap.
Amplify has issues that will screw you if you try to do anything with it that isn't a small project or like a personal blog. I have seen more than one business that got stuck between a rock and a hard place because their devs made an early decision to go Amplify vs. go pro.
yes, facing the same thing. can you explain what does pro mean? do you mean a manual cloudformation and cicd setup?
Any update on this caching issue?
Nope, I’ve moved past amplify and haven’t gone back
@@WebDevCody do you have a primary recommendation now? also by aws are you referring to amplify or everything
@@gamalf123 just amplify
SST received funding from YCombinator so should be around for a while.
awesome, thanks for pointing that out, that actually makes me trust it a lot more
I think amplify has actually been abandoned for the most part, or the management is just really bad. I have been having an issue with trying to deploy non static next3 apps to Amplify and I've been unable to for months, saw an open issue that has been there for almost a year, if not more. and it also had not response from the dev team at all. I ended up just doing it myself with lambda and cloudfront. When It works, it's great. when you have issues, there's no way of fixing or finding a fix for it cos nothing has been done to improve it in months. thinking maybe this could be a good project to do, light weight version of amplify with the things people actually want and no hidden resources cos I love the setup with appsync, dynammodb and every other thing that comes with it. with they were more passionate about Amplify or more active. It's one thing to not want to improve something, It's another thing to completely ignore the entire community for over a year. vercel is a no go for me, too expensive
I know this is 6 months old so you probably got an answer, but “user defined cost allocation tags” are what you want. :)
Thank you!
Don't mistake AWS fixing customer issues with what you see in the github repo. the two are pretty much completely disconnected. Probably the thing to do would be to get support tickets raised and linked to the github issues.
Use SST for everything. Don’t ever touch amplify with a 10 million foot pole. That crap is instant tech debt
Yep. Had an issue with their React library. I just cannot imagine how a team of first-class engineers who can be paid 4 times my salary can produce a crap like this. It was ~3 years ago I think, so maybe they changed it but it was painful at the moment.
@@danyfedorov3029 well, the folks that made SST feel the same was about the aws-cdk team lol
@@zheil9152 but they are still using CDK underneath, right?
Tbh Amazon really doesnt care about all their services, Amplify 100% was built in a reactionary way to compete with Vercel products, because of the attrition and project changing stuff like this doesnt get maintained as much as it should. You can even see this through documentation not being as robust as watching tutorials or other explanation videos. AWS is super useful but it really makes you feel like you are on your own.
I always hear ec2 makes aws the most money, so it’s hard for a company to invest time in their other products if they don’t make much. I would think amplify is trying to compete against google firebase
AWS will eventually screw you over if you have the wrong setup. Digitalocean I know what I'm paying for up front and my db and other services are all on DOCN. No surprises.
Maybe it’s a ‘Feature’ not a flaw? #chaching
maybe, I could just not understand the benefits for caching via cookies... 🤷♂️
Why would they address issues that would decrease the amount of money they’d get from customers lmao.
Because to the right customer you aren't talking cents youre talking thousands. That causes churn
AWS gets really expensive
Because.. thats how they farm money.
There are So many options to deploy to aws and they all sucks, cdk, amplify, app runner… etc not sure where to start
I just tried to setup a basic code deploy and elastic beanstalk and quit after 1 minute; it's all to complex.
some of thest things need to get standardized across cloud providers. like serverless framework. currently their proprietary solutions are not being well maintained
@@abeidiot totally its really annoying and frustrating
all for 1.65 USD a month or a day?
It wasn't one month or one day. You should try watching the video.
For maybe 7 days or so. The main point is it’s not caching my static files which is causing the price to be much higher than it should. If I had actual traffic to my site my bill would be much higher
@@WebDevCody totally understood your point, and you did a great analysis and you are through with the process too. I learnt a lot from this. I am planning to use AWS instead of Vercel. So this was right in time. Appreciate it. Do you have a discord?
@@gsanjeevkumar yes discord link in description
Google analytics is internet cancer anyway. I implore you to research some of the open soruce, privacy respecting alternatives out there
Yeah that’s true
GigaChad devs run their bare metal servers in their own datacenters. Big Iron for big men. None of this soyboi cloud (aka someone else's computer). You can easily replicated the shiny cloud dashboard by using something like Portainer. Mix that with Pxe Proxmox host, and you can instantly not only automatically expand your virtual compute but also underlying physical computer hardware, just shove another 1U server into the racks baby
Prompt source? I also want to troll other comment sections with a charicature of divorced 60yo republican C++ developer.
big iron for big men 🤣
@@Xevion chatGPT? nah bro, I was born in the free world outside of the Matrix in Zion. All my comments are natural and organic.
Why AWS will never be excellent? Because they don't use their own product.
I hate AWS.
It’s not too bad once you learn about 15 different services pretty well
Mind Juice
Amplify is a crazy mess and has caused many ruined nights for us. I've had too many breaking bugs prop up during builds with their default setup(they use the latest version of amplify cli automatically). It's more like a video game in terms of bugs than an enterprise cloud service