0

I'm experiencing a very weird behavior of the AWS Classic Load Balancer. From time to time, for no apparent reason, some requests to the load balancer randomly fail.

I saw 3 requests to the same url in the Chrome console. enter image description here

The first were successful: enter image description here

But the third failed. For some reason the "Remote Address" part is not there: enter image description here

The app keeps trying to send the request, it fails a few times and then it succeeds.

That happens only for some users for some time (never for the same users and never at the same time) and then it all goes back to normal. It happens for users in different networks and countries.

I checked the CLB logs and I can see the first two requests there but not the third. I contacted AWS support but they have no clue of what is happening. They asked me to run tcpdump on the machine next time it happens and send them the .cap file.

The fact that the "Remote Address" part is not there makes me think it is a DNS issue but I don't see how this makes sense given that it was working a few seconds/minutes before and it starts working again a few seconds/minutes later

  • I don't think you'll find this is ELB related. The browser would not show the request as "canceled" if the request had *failed*, so I suspect the server and balancer and even DNS are not involved. See https://stackoverflow.com/q/12009423/1695906 for some possibilities, and see what you find if you test with Firefox or IE. – Michael - sqlbot Aug 26 '18 at 23:37
  • Angular cancelled the request because it took longer than the timeout I set when I sent the request. I updated the 3 requests images in my post, take a look. The first 2 requests are done in ~100ms, the other one is cancelled after 12s (that was the timeout I set when doing the request using Angular's HttpClient). It looks like the request hangs somewhere, I don't know how to explain the fact that it never reaches the load balancer. – Felipe Taiarol Aug 27 '18 at 09:23

0 Answers0