i have a question about lambda anti patterns, and how to address my specific situation.
My current setup is this:
user/webpage -> ApiGateway -> Lambda1 -> synchronously calls Lambda2 (my other microservice) -> back to lambda1 -> back to user
Currently my lambda2 is behind an apigateway as well, but I toyed with idea of invoking directly. Either way it's basically another microservice that I control.
I understand that generally, lambdas calling other lambdas are considered an antipattern. All the blogs/threads/etc online mention using stepfunctions instead, or sqs, or something else.
In my situation, I don't see how I could use stepfunctions, since I have to return something to the webpage/user. If I used a stepfunction, it seems like I would have to then poll for the results, or maybe use websockets; basically in my webpage I would not be able to just call my endpoint and get a result, I'd have to have some way to asynchronously get my result.
Similarly with a queue, or any other solution I saw online, it's basically all asynchronous.
Is there any other pattern or way of doing this?
Thanks.
While invoking a lambda from another lambda, everything will work fine except when the second lambda timeouts or it throttles. If your business logic is built in such a way that failures are handled gracefully and has idempotent behaviour built in, then a lambda calling another lambda (via API gateway or direct invocation) should work fine. Having said that, AWS has come out with Synchronous Express Workflows for AWS Step Functions. The linked blog has detailed examples of using it. The only caveat here is that your entire operation should get over in 5 minutes. The maximum duration an express workflow can run is 5min. So if your application flow is completing within that time limit then this is the recommended way of orchestrating services.
Related
Im a little confused since AWS has a lot of features and I do not know what to use.
So, I was creating a Lambda function that does a lot of stuff with a remote web, process could last at least a minute.
So my idea was creating an API that calls this lambda, have lambda create an unique ID and return a response right away to the client with a token., save this token to a DB.
Then have lambda process all this stuff with a remote web and, when it finishes, save the results to the DB and a bucket (a file), so this result is ready to deliver when the client makes another call to another API that makes a query to the DB to know the status of this process.
Thing is, it seems that if a response is sent from the handler, lambda terminates the execution, Im afraid the processing with the remote web will never finish.
I have read that step functions is the way to go, but I cant figure out which service will take the processing, ¿maybe another lambda?
Is there another service that is more suitable for this type of work?, this process involves scrapping a page and downloading files, is written in python.
I have read that step functions is the way to go, but I cant figure
out which service will take the processing, ¿maybe another lambda?
Yes, another Lambda function would do the background processing. You could also just have the first Lambda invoke a second Lambda directly, without using Step Functions. Or the first Lambda could place the task info in an SQS queue that another Lambda polls.
Is there another service that is more suitable for this type of work?
Lambda functions are fine for this. You just need to split your task into multiple functions somehow, either by one Lambda calling the other directly, or by using Step Functions, or an SQS queue or something.
I have a use-case with lambda concurrent.
My system uses the API Gateway + Lambda function. I requested AWS for more Limit Concurrent and I get 10000 CCU. But my system is going to solve about 30000 concurrent users.
First, I tried using lambda async by this guide [https://docs.aws.amazon.com/apigateway/latest/developerguide/set-up-lambda-integration-async.html]
But my application use REST API and the result get 500 code of the response. The document said
"In this case, the backend Lambda function is invoked asynchronously,
and the front-end REST API method doesn't return the result"
Then, I search for one more solution that use Lambda Async. It used lambda async invoke from another sync Lambda. But this way didn't use for API GET of some methods that required validation.
Now, we go back with the solution that uses the provision concurrent, but this solution may not be useful. I am concerned about the concurrent that not enough.
Please, help me. Sorry about my English skill
I have defined a lambda function that is invoked from API Gateway with proxy integration. Thus, I have defined an eager resource path for it:
And referenced my lambda function:
My lambda is able to process request like GET /myresource, POST /myresource.
I have tried this strategy to keep it warm, described in acloudguru. It consists of setting up a CloudWatch event rule that invokes the lambda every 5 minutes to keep it warm. Unfortunately it isn't working.
This is the behaviour I have seen:
After some period, let's say 20 minutes, I call GET /myresource from API Gateway and it takes around 15 seconds. Subsequent requests last ~30ms. The CloudWatch event is making no difference...
Let's suppose another long period without calling the gateway. If I go to the Lambda console and invoke it directly (test button) it answers right away (less than 1ms) with a 404 (that's normal because my lambda expects GET /myresource or POST /myresource).
Immediately after this lambda console execution I call GET /myresource from API Gateway and it still takes ~20 seconds. That is to say, the function was still cold despite having being invoked from the Lambda console. This might explain why the CloudWatch event doesn't work since it calls the lambda without setting the method/resource-url.
So, how can I make this particular case with API Gateway with proxy integration + Lambda stay warm to prevent those slow first request?
As of now (2019-02-27) [1], A periodic CloudWatch event rule does not deterministically solve the cold start issue. But a periodic CloudWatch event rule will reduce the probability of cold starts.
The reason is it's upto the Lambda server to decide whether to use a new Lambda container instead of an existing container to process an incoming request. Some of the related details regarding how Lambda containers are reused is explained in [1]
In order to reduce the cold start time (not to reduce the number cold starts), can you try followings? 1. increasing the memory allocated to the function, 2. reduce the deployment package size (eg- remove unnecessary dependencies), and 3. use a language like NodeJS, Python instead of Java, .Net
[1]According to reinvent session, (39:50 at https://www.youtube.com/watch?v=QdzV04T_kec), the Lambda team expects to improve the VPC cold start latency in Lambda.
[2] https://aws.amazon.com/blogs/compute/container-reuse-in-lambda/
Denis is quite right about the non deterministic lambda behaviour regarding the number of containers hit by CloudWatch events. I'll follow his advice to improve the startup time.
On the other hand I have managed to make my CloudWatch events hit the lambda function properly, reducing (in many cases) the number of cold starts.
I just had to add an additional controller mapped to "/" with a hardcoded response:
#Controller("/")
class WarmUpController {
private val logger = LoggerFactory.getLogger(javaClass)
#Get
fun warmUp(): String {
logger.info("Warming up")
return """{"message" : "warming up"}"""
}
}
With this in place the default (/) invocation from CloudWatch does keep the container warm most of the time.
I currently have a 3rd party application pushing messages to a Lambda function through API gateway. The Lambda function needs to serialize, log, and push the message to another ESB that I have very little control over.
I'm trying to ensure that there is some kind of recovery mechanism in the case that the Lambda function is either at max load or cannot communicate with the ESB. I've read about Kinesis being a good option for exactly this, but the ESB does not support batching for my use case.
This would cause me to run into the scenario where some messages might make it to ESB, while others don't, which would ultimately cause the batch to fail. Then, when the batch is retried, the messages would be duplicated in the ESB.
Is there a way I could utilize the functionality that Kinesis offers without the batching? Is there another AWS offering that better fits my use case? Ideally I would have one message being handled by the Lambda function that stays in the queue until it is successfully pushed into the ESB.
Any tips would be much appreciated.
Thanks,
Matt
Following might be of help to you:
1) setup api-gateway to log to sqs and 2) then set up a lambda function on that sqs queue to serialize, log, and push the message to the external endpoint.
For the first part: How to integrate API Gateway with SQS this will be of help. (as already mentioned in comments)
This article might help you more for second part: https://dzone.com/articles/integrate-sqs-and-lambda-serverless-architecture-f
Note that you can also choose what kind of trigger you would like (based on usecase)- cron based poll/ or event based, you also have control over when you are deleting from sqs in your lambda function. (you can also find the very basic code in lambda blueprint with name "sqs-poller").
Thanks!
My Amazon Lambda function (in Python) is called when an object 123456 is created in S3's input_bucket, do a transformation in the object and saves it in output_bucket.
I would like to notify my main application if the request was successful or unsuccessful. For example, a POST http://myapp.com/successful/123456 if the processing is successful and http://myapp.com/unsuccessful/123456 if its not.
One solution I thought is to create a second Amazon Lambda function that is triggered by a put event in output_bucket, and it to do the successful POST request. This solves half of the problem because but I can't trigger the unsuccessful POST request.
Maybe AWS has a more elegant solution using a parameter in Lambda or a service that deals with these types of notifications. Any advice or point in the right direction will be greatly appreciated.
Few possible solutions which I see as elegant
Using SNS Topic: From your transformation lambda, trigger a SNS topic, with success/unsuccess message, where SNS will call a HTTP/HTTPS endpoint with message payload. The advantage here is, your transformation lambda is loosely coupled with endpoint trigger and only connected through messaging.
Using Lambda Step Functions:
You could arrange to run a Lambda function every time a new object is uploaded to an S3 bucket. This function can then kick off a state machine execution by calling StartExecution. The advantage in using step functions is that you can coordinate the components of your application as series of steps in a visual workflow.
I don't think there is any elegant AWS solution, unless you re-architect, something like your lambda sends message to SQS or some intermediatery messaging service with STATUS and then interemdeiatery invokes POST to your application.
If you still want to go with your way of solving, you might need to configure "DeadLetter queue" to do error handling in failure cases (note that use cases described here are not comprehensive, so need to make sure it covers your case) like described here.