AWS Loadbalancer issue



First of all we've been using your lib for quite some time and it's worked great and saved us alot of time, so thanks for all your hard work. Now, we are moving some of our sites to amazon and we have two machines behind an aws app load balancer. For some reason this causes an issue where bundles sometimes take a very long time to load. Requests to bundles can take more than 3 minutes, not always though. If we disable the browser cache we don't get this issue. Also if we bypass the load balancer and hit one of the webservers directly we don't get any issues.

Just wanted to ask if you have any idea what the reason for this could be?
Closed Mar 10 at 10:14 AM by Taritsyn


ZNS wrote Mar 6 at 9:08 AM

After researching this some more it seems the issue occurs (not always though) when the browser sends "if-modified-since"

Taritsyn wrote Mar 6 at 10:21 AM

Hello, ZNS!
If we disable the browser cache we don't get this issue.
Describe the configuration settings, that you use for this?

ZNS wrote Mar 6 at 12:07 PM

We are running the latest version of core, less and mircrosoftajax together with latest version of msiejavascriptengine. Pretty straight forward config. Here's the less

<less useNativeMinification="false" ieCompat="true" strictMath="false" strictUnits="false" dumpLineNumbers="None" javascriptEnabled="true" globalVariables="" modifyVariables=""><jsEngine name="MsieJsEngine" /></less>

I haven't excluded that this is some issue with the amazon load balancer. I'm looking into that now. But it's a bit strange that I don't get this issue with other content.

Taritsyn wrote Mar 6 at 12:14 PM

How do you disable the browser cache?

ZNS wrote Mar 6 at 1:11 PM

It's possible to do in most browsers under developer tools (F12). Anyway I've tracked the issue to the load balancer.

So the problem only occurs (but not every time) when both of these are fullfilled:
  1. The request sends "if-modified-since" in the header
  2. The request passes through the amazon load balancer (if we connect right away to the server there's no issue)
This does not make any sense to me, so I'm thinking this is an amazon issue.

Taritsyn wrote Mar 6 at 4:19 PM

It is unlikely I can help you in resolving this issue.

In any case, I recommend you to read the “Bundling behind cluster of cache servers” issue.