0

I have an application with about 200k users and am running a NGINX + Gunicorn(Python) server behind an AWS EC2 loadbalancer.

I don't understand how my requests is always 4k/minute but only sometimes I get half of traffic being timeout issues. Most of the time all requests are fine, but sometimes it starts to lock up then almost all requests get timeouts.

I noticed this pattern of the # of current connections has a wave and fluctuates from 1000s to 0. Is NGINX bundling requests somehow? How can I differentiate request_time to figure out whether it's NGINX not being configured properly, or my Python server just getting extra slow endpoints being called too often.

I've attached a screenshot of one of the servers in my NGINX Amplify dashboard.

Any ideas of parts of NGINX logs or Amplify that I can investigate to determine whether this is an NGINX configuration issue, or if the hosted Python process is getting locked up? Thank you! NGINX Amplify Dashboard

Devon Bernard
  • 101
  • 3

1 Answers1

0

Look in your CloudWatch metrics, especially for metrics about "dropped" or "failed". You can see all the details there about your loadbalancer and your EC2 instances. I don't know what are the instance types you're using but it could be that you're constantly overusing T2/T3 instances and run out of credit. I would think that some parts of your flow are being throttled for some reason and it's not neccessarly a problem on the instances themselves.

Chris
  • 318
  • 1
  • 5