Slow ssh connect

[bryanlharris]

Hi folks,

Before I start, I don't think this is the classic DNS issue (although I may not know enough to know that).

I have had an OpenBSD server for about a year now at a colo company. I'm using it to learn about OpenBSD things. All this time I have either connected via Windows 7 PuTTY or MacOS command line ssh. I have not had any issues doing it that way, and still don't.

The trouble is I have installed the default RHEL 7 on my iMac at home and the ssh connection to my OpenBSD server has that initial pause thing. I checked on the OpenBSD side and the UseDNS=no is in place (because it's the default).

I realize it is very strange for me to ask an obvious RHEL 7 question here on the OpenBSD forums. I just thought perhaps it might be something obvious that someone has encountered before.

A google search shows me all the UseDNS=no stuff, and when I eliminate that I don't find an obviously useful direction to go in.

I had a few ideas, like my RHEL 7 client might be trying all that GSSAPI stuff. I could disable that. Or it could be that my ed25519 key is somehow related to it. I could try using an RSA key. I just thought I'd ask first before trying random stuff.

Does anybody have any experience with this specific type of problem? I.e. only one O/S seems to connect with the annoying slow-pause, others connect and there is no pause.

V/r,
Bryan

[bryanlharris]

To clarify the thing that's bugging me. When I login via ssh (but only from RHEL 7), the initial ssh connection has a very annoying pause. Everything works, eventually, after the initial pause.

[jggimi]

Try connecting with one, two or three repetitions of -v, which may be helpful. See ssh(1).

[J65nko]

The delay is caused by a failed reverse DNS lookup issue ;-)

There is SSH setting called "UseDNS" that controls whether this reverse DNS lookup should be done.

From sshd_config(5):

 
UseDNS
    Specifies whether sshd(8) should look up the remote host name, and to check that the resolved host name for the remote IP address maps back to the very same IP address.
    If this option is set to no (the default) then only addresses and not host names may be used in ~/.ssh/authorized_keys from and sshd_config Match Host directives.

A few years ago this was set to 'no' on OpenBSD. Looks like it is still enabled on your RedHAT server.

You could run tcpdump on the server to monitor port 53 (named of DNS) UDP as well as TCP traffic. Then you can see the issue very clearly