Slow SSH Establishment

[Martin2]

I have Bell Fibre Internet into a new condo in a new building in Toronto.  For work, I use command-line SSH to tunnel into Linux servers.  The initial connection establishment typically takes 8 to 10 seconds (it seems to hang).  Once the connection is established the speed is great.

Has anyone on here experienced the same problem and have a solution?

I phoned Bell about the issue and they tried changing my DNS servers, which never helped.  They also sent me a brand new modem (but the other modem was brand new too), and that never helped.  They told me their engineering team was aware of the issue and would be fixed in 2 weeks.  That was over 2 months ago.  I phoned Bell again recently and they escalated me to what I was told is the "highest level technical support team", to which I had to explain the entire problem again to them, and they told me that they had never heard of "SSH" and that they don't support this type of technical help request.

[Vanadiel · Accepted solution]

Check your SSH config to see if it requires reverse DNS.

Most Bell routing points do not have the configuration for both forward(A) and reverse(PTR) records, leading to this issue with SSH if you have the (default) option enabled to perform a forward and reverse DNS record check.

You can check this for yourself when doing a traceroute to the remote SSH server. You will see the IP's of various Bell routing points, and usually the first 2 or so do not have a DNS record associated with them. The ones after that will have a DNS record associated with them.

I am thinking your initial delay has to do with the inability to perform the forward and reverse DNS record check.

[Martin2]

I'm on Mac book and in my /etc/ssh/sshd_config I have "UseDNS no".  I'm not sure what steps I should take other than that.  

[Vanadiel]

That should be about it as far as I know.

[leb1]

I have  the same issue with ssh.  Over Bell Fibe takes at least 10 seconds. Let me know if anybody figures this out.

[Martin2]

Hi leb1,

If you have access to the sshd_config on your server (/etc/sshd/sshd_config) ensure you have set "UseDNS no"

This worked for me.  This is still a Bell issue though.  We shouldn't have to modify server settings just because an ISP is doing something unconventional and they won't fix it.  I've been ssh'ing for three decades into Linux servers and never once had an issue until I used Bell Fibe at home.  Whatever it is Bell is doing (I don't care to know the technical aspects, I'm busy with my own stuff), they shouldn't be doing.

Happy Easter

[Vanadiel]

It's because the first 1 or 2 routing points do not have RDNS records. SSH default config does a forward and reverse DNS lookup, which will fail.

It's a common issue.

[leb1]

I've been ssh'ing for decades too and never faced this issue till Bell Fibe. Rogers is fine.

But good news is Bell appears to have fixed the issue. Just tested several ssh connections to Linux servers and logon lightening fast now. I  made no changes server side (as recommended by Martin1).

[sman]

Just switched to Bell from rogers and noticed this exact issue on one of my server.  I had to add UseDNS NO to the server to make ssh connection not take 30 seconds.

Weird that this had to be configured on one of the server while the other servers were fine.

Never had I had any SSH issue with Rogers.