CODEWITHSUNDEEP
facebooknode.jsamazon-web-servicesnginxamazon-elastic-beanstalk
CargoMeister
CargoMeister

Reputation: 4299

Problems with AWS Load Balancer Timeout and No Error in Node/Express

We have run into the infamous AWS Elastic Load Balancer timeout of 60 seconds. We are interfacing to various social media engines. Sometimes the calls take longer than 60 seconds. One was definitely an error, but we got no error in our logs, only the 499 code in the nginx log indicating that the client, (the load balancer,) had severed the connection.

Anyone have some great idea for how to interrupt and grab a stack trace when we reach a certain length of time in the call? This is incredibly frustrating. We found one error, only because it was in one of our developers accounts, so easy to run on system with no ELB, and voila, the error occurred and was understandable. But this could be caused by one problem, or many problems, or simply that Facebook is having a bad day. But it would be nice to intercept what is happening to troubleshoot.

Upvotes: 1

Views: 546

Answers (1)

Rico
Rico

Reputation: 61531

You can try using GSLB. Who is your DNS provider? Ask them if they provide GSLB.

Here's a bit of an explanation: http://www.a10networks.com/products/axseries-gslb.php

In essence, you can setup different access points to your applications whether it is a load balancer or a single server and you can distribute the traffic according to weights. You can play around with the weights, send some traffic to your single server application and see what's going on.

Upvotes: 1

Related Questions