Deploy FastAPI on AWS Lambda ⚡ Serverless hosting!

Key Takeaways

hey everyone welcome to this tutorial where you'll learn how to host a fast api application serverlessly on aws lambda in my previous video we hosted fast api on amazon ec2 which you can think of as virtual servers but some of you asked how to get it production ready meaning how to use it as part of a real business or service honestly with ez2 that's quite a challenge you have to think about things like how to scale out your hosts when your traffic increases how to load balance the traffic how to do rolling updates for an app and how to do os and security patching not to mention it'll be expensive to keep running because you'll be paying for uptime of the server whether you have people using your api or not that is why i prefer to use serverless compute for hosting apis and on aws the serverless compute offering is called aws lambda we can create something called a lambda function which is basically a piece of code we want to run in the cloud except we let aws handle all of the scaling security hosting and load balancing problems it scales very well it's also cheap because we only pay for what we use but the most important thing is that it takes a lot of the boring work off of our plate so how do we use it with fast api well in this video we're going to take a fast api application we're going to modify it to work with aws lambda we're going to upload it test it and then turn it into a live http endpoint and unlike in our ec2 tutorial this api is going to be more or less production ready right out of the box let's get started to work on this project you're going to need to have a fast api project or application ready to use for this one i'm going to be using a project i created as part of a previous video so you can follow along with that if you don't have a project check in the comments below for a github link and you can just clone it directly you also need an aws account and have some knowledge with lambda functions i have a video you can check out on that as well if you're not up to speed with that yet so once you have those two things set up you can start here i've opened up the project folder in my editor which is vs code and the project's really simple there's only one file which is the fast api application file if you're using your own project then you might have a file as well or you might have multiple files it doesn't matter the important thing we're going to look at is this line which we use to create the fast api application so all of this path stuff and all this app stuff that we have here it doesn't we're not going to change any of that we're just going to focus on this thing the problem with running fast api on aws lambda is that lambda functions don't understand this api interface this app get app you know with the routes and everything this agsi interface is not something that's understood by lambda functions instead lambda understands something called a handler function so a basic handler function looks something like this it's basically just a function that takes in two arguments an event and some context and these are both dictionaries when we create a python file that we want to run as a lambda function we have to tell it what we want it to consider the handler function and when that lambda is run it will create this python runtime with these files and it will run this function that we specify for it and then it's going to pass in anything it thinks is relevant so when we hook up a lambda function to an http endpoint maybe with api gateway or whatever technology we use all the things like the path information the headers all that stuff is going to be hidden away in these event and context objects and so we have to find a way to unpack that data and have it route to the appropriate fast api path or this sub function here that we have now we can go the long way by unpacking this ourselves and then figuring out which of these functions to use depending on what event is given or we can sort of take a shortcut and use an existing solution that someone already came up for this and we can use this library called mangum which is an adapter for running agsi applications that's our fast api application there in aws lambda to handle function url api gateway etc etc so the way we use it is we just first have to install it so go to your python environment there and just type pip install mangum and once that's done you can wrap your entire app with a mangum handler so let's go and have a look at their documentation and it's pretty much basically just this first two lines here we have to import it first and now we'll create our handler using that wrapper and i can delete this handler here so now this handler function is something that aws lambda can understand and it's going to route that function request correctly to one of these sub paths so just save that file now our file's ready we need to upload it to lambda so let's go ahead and try that and see what happens go into your aws console and search or click on lambda and once you're there click on create function to make a new lambda function and we're just going to do it from scratch and you can call the function whatever you want i'm going to call it fast api bookstore and then for the runtime we'll choose our python 3.9 or whichever one is the one you want and then we don't really need to do any of this other stuff we can just leave everything else at its default setting so just click create function and wait a while for that to create once it's finished you should see something like this there's a couple of different elements here but we don't have to worry about this top part and if you scroll down a little bit you see the code source so this is what i was talking about earlier where this is going to be the code the python code that your function will run when this lambda function is invoked and here they've written some kind of hello world code for us which just a lambda handler that returns 200. so we can actually test this out now before we change anything just to make sure it works and we're going to need a new test event so we'll just call it hello and then this is the event object that's going to go into the lambda function we don't need to change that we'll just save and then we'll run it and there you go this function actually executes and it returns this hello from lambda text there okay so very basic let's see what happens when we just copy our fast api application and just put it in here so i'm literally just going to copy the entire code and i'm going to put it here but to keep it consistent with our development environment i'm also going to rename this to main.py okay so now that's ready and this function uh is pretty much the same as we have locally now if we actually want to save this change we have to click deploy because here it says change is not deployed so it's if we run it again it's still returning the previous version we had so let's go ahead and hit deploy and once this banner turns green it says it's successfully updated so we hit test again and this time we have an error because we can't find the lambda function module in this file that's because we changed the name of the file to main.py and we also changed the handler name from lambda handler to just handler on its own so we have to tell this lambda function where to look for this this handler function that we created so if you scroll down a little bit you'll see this runtime settings here and here is the handler setting which it's looking for a file called lambda function and a function called lambdahandler in that function so we actually have to edit this and then we're going to change this to main because that's what our file is called and then our function is called handler so main.handler then save okay so we don't actually have to deploy anything again because the code itself didn't change and their settings that we changed below took effect here as we can see so let's go ahead and run this again and see if it works okay so this time it's a different error it actually found the handler function but now it fails because it can't import fast api and this is actually a very common type of failure for people just getting started importing their python functions into lambda because you actually also have to have all the dependencies when you're running a function in lambda it actually this is the only thing it has in its runtime it's got the python library i think it might have the aws api library as well built in by default but apart from that the environment has nothing else so things like fast api or anything that you install with pip you also have to find a way to put it inside this folder here now that doesn't actually sound like a very bad problem like let's just drag and drop all our dependencies in there right but actually it gets more complicated than that for example what if you want to be able to update them how do you keep this function updated how do you keep the dependencies updated and how do you do it in such a way that you can have it working in your dev environment and also synced up to your lambda function in the cloud there there's actually a lot of different ways this problem can be solved if you're familiar with docker then lambda is also able to run docker so you can use that to install all your dependencies there and then just run that docker image you can create something called a lambda base layer where you install all your dependencies once and then it's already existing for anything in your function that uses that base layer or you can pretty much zip up the entire dependency directory on your disk and then upload it to the lambda function and because that's the most straightforward way and the easiest thing to do that's what we're going to do here in this tutorial so now i'm back in my project folder and you can see that the dependencies are not anywhere here in the folder that's because i'm using a virtual environment that actually saves it to a different location so it's not here and you might be doing the same as well or you might be using the libraries directly in your base python interpreter so we have to find a way to install those dependencies into a local directory now if you go back to the readme in this github project this fast api aws lambda github project i have in the comments i've actually pasted a few useful commands that we can use to do all those things so if we run pip install with a t flag and a folder name it's going to actually install the files or the dependencies we want into that subdirectory so i'm going to go ahead and do that and i'm also going to install it from a requirements.txt rather than individually although you can do that as well so first i'm going to add mangum the library that we just installed to our requirements.txt and then in my terminal i'm going to type pip install t libraries which is going to be standing for the libraries of this project and then r requirements.txt and once that's done if you go back to your project you should see this new lib directory created with basically all of the python dependencies we need for this project so we're going to have to zip all of that up and upload it to lambda but because this is installed as sort of like a transitory library we're not actually using this in our project i'm going to want to add it to my git ignore so that i don't accidentally commit this and just have a bunch of useless code going around with me in my project so we're just gonna add lib to that and that should just turn gray okay so now how do we zip this up when i zip this up i actually want all of these files all of these folders expanded in the root path so even though i've put them in this lib folder here when we have it in our lambda function for our main.py to be able to access it it can't be in the lib folder it should just be in the space directory so if you go back to this readme the next command will tell you how to do that we'll basically have to change directory into the library file and then zip it there but then we're going to zip it into a file called lambda function in the sort of the previous the parent directory and this dash r means recursive so we zip everything in every file there as well and the dot just means do it in that directory so let's go ahead and actually run this in our terminal okay once that's finished running i'll go back to my project again and i can see that there's a new lambdafunction.zip file here now i'm also going to add the zip file to our ignore list because i don't want to commit that one either and now i actually the only thing i have to do now is just add this main.py to that zip file so again back to the readme there is one last command we have to run it's zip lam the function.zip dash u and then main.py so the dash u says that don't replace the zip file just update it or add new things to it and this is the file we want so just go ahead and copy that one so zip lambda function dash u and then our main.py file okay and now our api if you followed the last tutorial it actually needs a book.json file locally to act as its database storage which this is actually not a good way to do it with aws lambda because there's no concept of hard disk on lambda every time the function runs you can pretty much think of it as running a whole new computer so anything we save or write to disk is pretty much not going to exist after that function terminates maybe for 15 minutes while the lambda is still warm which is how long a function stays alive for on that machine before lambda kind of recycles it and creates a new one but the bottom line is we can't really write or read files from there and expect it to persist uh it's just that that's not a problem i'm trying to address in this particular tutorial so we're just going to ignore it for now i'm just kind of letting you know ahead of time for the time being this books.json will work nicely with our main py to provide a data source for our api so let's go ahead and add that to our zip as well okay so now i finished adding those files to my zip and if i open the zip folder up in my browser i can actually see all of the dependencies along with my main.py file and mybooks.json data file so now this zip file is ready to go if i put this on lambda i can just run it so let's go back to our lambda function in the aws console and in our code source go to this button that says upload from and then here you can choose a zip file so click on that and then select your zip file to upload and hit save and now when it's finished you're gonna see that the code source doesn't actually let us look or modify the package contents anymore it's just a zip file at this point but that's okay because we already know what's in it from when we built it locally now let's go ahead and run the function again and see what happens this one still fails but if you open up the log you're going to find that it's a different error this time it says the adapter was unable to infer a handler to use for the event this is likely related to how the lambda function was invoked are you testing locally make sure your request payload is valid for a supported handler uh so translated into english this basically means the payload we're giving like the stuff we're giving to lambda this stuff is not something it understands it can't parse this it doesn't understand the http path the routes the headers uh it's expecting that information but it's not there so what we can actually do so that we can test using this interface is create a new event and then if you go to the template there's one called api gateway aws proxy which this is actually the kind of information that we're going to get when we hook up the lambda function to an actual api endpoint but this is a template so we're going to have to modify a couple of things first of all we're going to probably have to pick a path to use and an http method to use as well i want to do something really simple so i want to just pick this um maybe this root path so i'm going to turn the path into just my root slash directory and then i'm going to make it a get request so let's go ahead and modify this payload to do that now the payload structure so that this information is actually repeated several points throughout the json uh data so you're just gonna have to go through and change each of the times you see those fields and don't worry if you don't understand all the other stuff that's going on here you don't have to touch any of that at this point so now that's been modified i'm just going to save this event oh i have to come up with a new name for it i'm going to call it fast api http payload hit save now i'm going to test it again this time it's successful and if we go to the details you can see that the response is welcome to my bookstore app so there you can actually see that it's successfully figured out that it needs to call this function and it's returned this message from us so now our fast api is working on aws lambda i can also change it to a different path so for example maybe list books i'll copy that over here so let's say we wanted to test that we can change our event and we'll just make our path list books but we have to change it everywhere it appears i think it appears three places in total in this payload okay and then we hit save then test again and now you can actually see it list all the books that we've bundled up in our json folder so far we're doing pretty well we have a fast api application now wrapped up in a handler and we've put it into a lambda function so we've tested it and it works we can run this and we don't even need to have a server to run it and it's in the cloud the last thing we need to finish this project off is to actually make a public http endpoint so that we can actually call this api anywhere with http connection so from a browser from another app from a front end so let's get that working now this actually used to be a pretty complicated thing to do before we would probably have to use another aws service called api gateway and we'd have to configure that and make it connect to the lambda with something called like a proxy test those interactions separately and make sure that works but today there's actually a new feature in lambda that makes it really really easy to create an http endpoint right out of the box if you go to configuration there's a tab here called function url and you really don't have to do anything you just hit create function url and here we can select off type none and then configure cross domain origin access just leave that as the star and pretty much leave everything else as default and it's gonna just give you a bunch of other information you don't have to pay attention for now you just hit save and now we are done we already have and a function url that we can use so let's go ahead and check that out in our browser so i've opened up that function url and i'm going to zoom in and you can see that it's a public url that responds to this so you can actually view this from anywhere try it on a phone or a different computer you can actually access that here some of you might be wondering how do i get rid of this ugly api url and have my own domain name in that case i'm not quite sure you might still have to go with api gateway and then hook it up with route 53 in order to get that dns resolution but if your goal is just to get an api that you can use internally or as part of a bigger app or service then this api is good to go it's public and it's behind an https protocol right out of the box and we can test this further by actually typing in routes like list books and you can see all the books that we have here in that json file we can also try some of the other routes we have there like a random book and that one works as well and now our api is ready to go but do you remember what i said before about this json file and this lambda function sort of not having a persistent state that's actually kind of important because lambda doesn't keep its server like when i call this api what happens behind the scenes is that lambda says hey we're receiving a request to run this function let's provision a virtual server to run this and on that virtual server it's going to copy all of this code that we had including that json file and then we can run it we can read from the file we can even write to it but as soon as that function's finished that environment will no longer exist it's going to be recycled by lambda so anything that we do to change that will be lost and the next time the function is run it's going to kind of be fresh from here again so if you're using a lambda function you can't rely on something like this saving to a local disk even though i kind of ignored that and just used it in this project what you actually want is to have this connect to a database amazon actually has a lot of databases that work really well with lambda but you can also use it with anything you're familiar with like postgres or sql or anything like that so you're going to need to have some other separate mechanism to store your data and you can't really rely on the local drive to do that and that's pretty much it on how to deploy a fast api application to aws lambda i hope you found this tutorial useful and if you have any questions or comments please let me know in the comments below if you need help at any point then check in my video description for links and project source code anyways i hope you found this useful and thank you for watching