[TriLUG] TCP connections taking forever to form?

[Steven Tardy via TriLUG]

On Tue, Feb 2, 2021 at 10:34 AM Brian via TriLUG <trilug at trilug.org> wrote:

> Hi Gang,
>
> Something very strange is going on with my Centos 7.4 system, which acts
> as my firewall and server of various things.  Mundane operations are
> taking ages to complete and I think I have narrowed it down to a problem
> establishing TCP connections.
>
> I finally tracked down the slowdown by combing through a particular
> shell-script I was using, so my examples below come from that (getssl).
>
> I'm testing with this command:
> $ time curl --user-agent "getssl/2.33" -I
> "https://acme-staging-v02.api.letsencrypt.org/"
>
> Running that on my desktop machine (which is downstream of the server)
> gives me a response right away:
> [...]
> real    0m0.375s
> user    0m0.016s
> sys     0m0.015s
>
>
> Running it on the server itself takes AGES:
> [...]
> real    2m7.775s
> user    0m0.076s
> sys     0m0.094s
>
> Two MINUTES for a simple http request!  How can I figure out what in the
> world is causing this?  I'm stumped...
>
> TIA,
> -Brian


I did the same on a lab VM which (I didn’t think about at the time) does
NOT have internet/NAT connectivity. . . . And it failed at 2m7.528s with
“Network is unreachable”. Does stdout show an error.

Like others said ‘tcpdump -nn port 53 or port 80 or port 443` may help
uncover what is happening (not happening).