i have a domain.com point to digital ocean loadbalancer ip to distribute to 2server. when i try to use access the web via domain.com it is extremely slow and sometimes even timeout. but when i try via my server hostname it is working normally, what possible cause to this? is the load balancer slow in forwarding request?
looks like problem solved by changing do load balancer setting and update it. maybe this make loadbalancer to restart and update.
Its an old case of mine, because i don't get any solution, so i build own loadbalancer using DO droplet and floating ip.
Looks like the haproxy setting is not suitable. On my case, i guess that my user has over the limit of max connection. So it cause the new connection is in queue and not serve directly. And because i can't change that setting so i build myself own loadbalancer and increase max connection setting from the default one.
Hope this help.
I have the exact same problem with a LoadBalancer I created through kubernetes. Port 80 http traffic is fast but port 443 ssl traffic is extremely slow. When I flip a setting like algorithm it becomes fast on ssl but then after a while reverts back to being slow.
Looks like I have the same issue.
80 http traffic is always fast. But according to webpagetest.org SSL Negotiation takes about 30 seconds. But when I click SAVE in Proxy Protocol on Load Balancer Setting page, SSL Negotiation becomes fast — 128 ms! But after a while it backs to being slow. Clicking on SAVE makes it go faster again for some time.
Is it a bug?
In my case I eventually realised that the IP provisioned for my loadbalacer was a victim of an ongoing ddos attack. I reprovisioned to get a new IP and after that it worked.
I setup a nginx ingress in my k8s cluster and saw the nginx log being spammed with a lot of strange HTTPS requests. Basically they were sending HTTPS requests but with binary garbage so at first I did not understand what it was. All requests origninated from the same 2-3 countries.