Lambda timeout on sqs sendmessage

Question

I'm sending messages to a queue from a lambda function. But some times sqs.sendMessage dont return anything and the lambda get timeout. This happens sometimes

I tried changing de code many times, to work with await, promises and callback but the error persist.

Lambda log

const sqs = new aws.SQS({apiVersion: '2012-11-05'});

//TODO: Validar campos obrigatórios nas mensagens de acordo com o tipo de mensagem
exports.sendMessage =  async (message) => {
  let params = {
    MessageBody: JSON.stringify(message),
    QueueUrl: 'https://sqs.us-east-1.amazonaws.com/....',
  };

  try {
    await sqs.sendMessage(params).promise();
    return {statusCode: 200, body: {data: "Notification sent successfully"}};
  } catch (e) {
    return {statusCode: 400, body: {data: e}};
  }
}

Fischer

What is the memory size configured in your Lambda function? Code looks good. – Zhdanov

Hello, 128mb. In CloudWatch logs the most expensive use is 80mb. – Fischer

The problem may be it. The more memory you configure, the better the configuration for the machine your code runs on is. For example, at 1792MB you get a dedicated core. Increase it to 1024MB and test it. – Zhdanov

I tried to increase it, but the problem persist. It looks like a problem with the aws sdk – Fischer

I doubt it. I have been using it for years and never ran into issues. It’s definitely something to do with your code – Zhdanov

A changed my code and set httpOptions: {connectTimeout: 5000} on create the instance of sqs. Now my lambda dont get timeout, but it get 5s to finish. Second aws docs this params is:Sets the socket to timeout after failing to establish a connection with the server. So the problem is the connetion with the server. – Fischer

I have seen something that seems very similar but with the python3.6 lambda runtime. In addition to sqs, it also happens with the Python Requests module. So far, I have come to the same concussion; that it's a problem with socket connection to the server. I have been unable to reproduce this behavior with requests to the same server when running the client code outside of lambda. – Ankle

@Ankle I'm curious if your lambda was VPC attached or not? – Dehiscent

@Dehiscent I believe it was but I don't remember this situation 100%. I have implemented request retries so the problem hasn't been bothering me anymore. – Ankle

Answer 1

I just had the same issue. The problem was that the lambda had 3 subnets assigned to it and only 1 of them had a routing entry to 0.0.0.0/0 Even tho 2/3 of the subnets were missing that routing entry, only 1-3% of the requests randomly timed out, so make sure to check that!

Intercommunicate

Answer 2

I had a similar issues. What I did to make it work was add in the endpoint url into the boto3 client call.

For Example: boto3.client('sqs',endpoint_url='https://YourVPCDNSEndpointforSQS')

Mulciber

Try formatting your answer to be a little more readable. Cheers – Lyell