claudia update seems to basically require node_modules to have all your modules in the lambda zip file, which effectively renders tree-shaking useless
--use-local-dependencies flag. That will use your local dependencies instead of doing the re-install.Also, if you have tree shaking, you probably use Webpack or something similar. In that case, you don’t want to deploy the source. You want to deploy the build instead. Claudia requires the package.json file and Claudia config. I guess you can copy a package.json file without dependencies and Claudia config file to your build folder before the deployment, and then run the Claudia update command.
--config claudia-prod.json option to do that).
I have a lambda that uses a layer (sharp) that requires specific Linux binaries for lambda.
What's the best way to deploy this using claudiajs?
I'm using TypeScript and during local development I have the sharp library as a devDependency.
When it's running in AWS, it'll resolve
import sharp from 'sharp'from the layer and when locally, from the devDependency.
But if I try to deploy it fails because it cannot find sharp.
In the docs it talks about using --use-local-dependencies for a similar use case where binaries are needed for a different platform but I can't see how that works with layers where that dependency will never be directly within the lambda's own node_modules
I have a process triggered by an HTTP request to API Gateway that can last longer than 30s. It's retrieving a bunch of data, doing some conversions zipping up and returning.
To work around the 30s limit of API Gateway I'm creating a presigned URL for the file which I want to return immediately in the API response.
Then go ahead and actually build the contents of that file.
So I do something like this
api.get(
'/searchcsv',
async (request: ApiRequest) => {
const filename = `some-file-name-${new Date().valueOf()}.zip`;
const url = await getSignedUrlforCsv(filename);
searchInventory(request, filename);
return {
url,
};
},What seems to happen is that the lambda execution is terminated as soon as the return occurs. Is this something to do with the way claudia-api-builder works or is this fundamental to API Gateway? It seems unlikely it's the latter as that would mean tightly coupling API GW and Lambda but I could be wrong.
You can’t continue Lambda execution after returning a result to the API Gateway, that’s not Claudia API Builder limitation, it’s the AWS feature. However, you can send a background message using the SNS notification (or SQS, EventBridge, etc.) and continue running in the background.
If you are generating the file, it would be similar to the following:
- HTTP request sends the SNS message, and then returns a presigned URL
- Frontend pings that S3 URL each second to see if it’s ready
- The SNS message triggers another Lambda function that does the logic in the background and saves the result to the S3 file.
sharp, this is the problem that Claudia doesn't solve really well, because it's not easy to solve it for all cases. Your solution works fine. In the Lambda function itself, sharp will be available like any other globally available module, so you can install sharp globally and then you should be able to pack your node modules without sharp using Claudia, and everything will still work after you deploy it.
You can’t continue Lambda execution after returning a result to the API Gateway, that’s not Claudia API Builder limitation, it’s the AWS feature. However, you can send a background message using the SNS notification (or SQS, EventBridge, etc.) and continue running in the background.
If you are generating the file, it would be similar to the following:
- HTTP request sends the SNS message, and then returns a presigned URL
- Frontend pings that S3 URL each second to see if it’s ready
- The SNS message triggers another Lambda function that does the logic in the background and saves the result to the S3 file.
Thanks @stojanovic - that's almost exactly the pattern I went with, except using SQS instead of SNS
- The endpoint creates the S3 file and returns a presigned URL, and writes a msg (basically the original request + filename) to an SQS queue
- A separate lambda has the SQS queue as a trigger and processes the data, and replaces the S3 file (using same key)
- The frontend polls for a file using the presigned URL with content
Hello guys, when run:
claudia update --use-local-dependencies
Happens:npm ERR! code EEXIST
npm ERR! path /private/var/folders/gp/tksqwx0x3w96hnd2p_zczk480000gp/T/tKqD1w/quryappgwaws-1.0.0-HAkaXS/package/node_modules/claudia/node_modules/aws-sdk/node_modules/.bin/uuid
I am following this guide:
Building a Serverless Application in AWS using React and Claudiajs
