boto3 client for lambda stops reading from a connection after 300 seconds

I have a Lambda function that contains code that needs to execute for 5 minutes or longer. I invoke this Lambda function using boto3 and wait for the response from the Lambda function (response being a json object).

I noticed that on using botocore.config.Config to increase read_timeout for the boto3.client the client keeps the connection open but stops reading from the connection after 300 seconds (5 minutes) and throws a ReadTimeoutError exception after the specified value of the read_timeout parameter.

Below are the code snippets to reproduce the issue:

Client machine code: python: 3.5 OS: Ubuntu 16.04 Linux botocore version: 1.14.11

import json

import boto3
import botocore

config = botocore.config.Config(read_timeout=310, connect_timeout=310, retries={'max_attempts': 0})
client = boto3.client('lambda', config=config)

payload = {"body":some_string}
payload = json.dumps(payload)

response = client.invoke(
            FunctionName='arn:aws:lambda:<region>:xxxxxxxxxxxx:function:function_name',
            InvocationType='RequestResponse',
            Payload=payload
        )

Lambda function code: python: 3.7

import time
import json

def lambda_handler(event, context):
    time.sleep(300)
    return {
        'statusCode': 200,
        'body': json.dumps('Hello from Lambda!')
    }

The above code produces the following error message: botocore.exceptions.ReadTimeoutError: Read timeout on endpoint URL after 310 seconds. I have tried experimenting with multiple values of time.sleep() and read_timeout. After referring to the CloudWatch logs I am fairly certain this only happens if Lambda executes for 300 seconds or more (the same can be verified by changing the sleep time in Lambda to 299)

I have referred to the issue #1104 and #205. What seems to be the issue here? Any help would be much appreciated.

[arcadecoffee]

We’re experiencing an issue exactly like what is described above. Was there ever any real resolution here?

[swetashre]

Closing this issue due to inactivity. Please reopen if you have any questions.