Just built a brand new virtual deskop using Vagrant & Puppet to automate the install of a CentOS virtual desktop. My goal is to build automation into the project from the ground up - the only way to truly do this is to scrap and rebuild many times yet I was having a big issue w/ a fundamental requirement.
I needed openconnect installed to VPN into multiple infrastructure nodes. Unfortunately, I wasn't able to connect to other infrastructure servers (via ssh) and the system was responding w/ a timing out exception.
To gain a better view I enabled verbose ssh output using the "-vvv" parameter and quickly saw the system was stuck at the "expecting SSH2_MSG_KEX_ECDH_REPLY" phase.
I was fortunate to have a working VM and performed multiple compare & contrasts against settings. With debugging enabled I was able to see the server was sending a host key over to the client. On the broken system it was not receiving that key...
After performing a ton of research to resolve the issue I found many mixed results from individuals. Seems there are a mix of server (router, sshd, etc) and client side configurations that could be changed to optimize the maximum transmission unit value.
Seems lowering the MTU value in the systems: /etc/sysconfig/network-scripts/ifcfg-enp0s3 worked.
It was timing out @ 1500 yet worked @ 1200
bash-4.2$ cat /etc/sysconfig/network-scripts/ifcfg-enp0s3