Hi Brendon, I just wanted to reach out and say that we found this too!
It seems that the Kernel developers decided to halve the default TCP memory in the 4.x kernels, it probably makes sense for most applications but not when dealing with busy high network usage like we typically see when acting as a load balancer and/or reverse proxy. The actual change is mentioned here: https://git.kernel.org/pub/scm/linux/kernel/git/davem/net.git/commit/?id=b66e91ccbc34ebd5a2f90f9e1bc1597e2924a500 For me reducing it by 50% didn't work well... So I wrote a script to simply double TCP memory if a newer Kernel is detected as I knew it was reduced by 50% from what I had been used to and it always worked for me on the old defaults. However, your method is better(Less lazy)... Simply decide if you need to increase it by looking out for the error message: TCP: out of memory -- consider tuning tcp_mem Anyway, just thought I'd mention it for info and to say you are not alone ;) Aaron West Loadbalancer.org Ltd. www.loadbalancer.org
