www/obhttpd and question marks in sockstat

[Niklaas Baudet von Gersdorff]

Matthew Seaman [2016-03-29 22:59 +0100] :

> Three hours does seem a little excessive for closing down a listening
> socket.  Yeah, that's not right.  Those already-closed connections
> should disappear after a few seconds.  Actually, it's also quite odd now
> I come to think of it, to see them for the listening socket of a server
> process.  It's common for them to appear for the point to point link
> when a particular client connects and then disconnects -- you can see
> the effect in sockstat on either the server or the client machines.
> 
> This is possibly related to the other odd output you've been seeing.
> Looks like the obhttpd process is not getting shutdown properly; rather
> it seems to be being killed in an excessively brutal way, which prevents
> it from cleaning up after itself.

Sounds reasonable.

> It might be worth trying some alternate web servers.  If you want a
> pretty bare-bones but fast alternative, I like www/nginx-devel with most
> of the port's OPTIONS turned off.
> 
> I know that nginx doesn't normally suffer from anything like this
> effect. If you still see it with nginx, then you've good evidence it's
> not something to do with a specific httpd implementation, but something
> systemic in your server setup.
> 
> Do you see anything similar for any other network servers on your
> machine?  Assuming, that is, you can afford to turn them off and
> on-again a few times while checking.

I've already installed www/nginx, started/stopped it, and it works, that is, no
"open" ports once I stopped it. So I'm sure that mail/obhttpd is malfunctioning
has something to do with mail/obhttpd and not the system.

Priyadarshan [2016-03-30 06:59 +0000] :

> As alternative to obhttpd, we have been using Caddy with good results.
> 
> It is not on Freshports (yet) but one could get most configurations for
> FreeBSD here: https://caddyserver.com/download

Wow. Looks interesting. Probably I'll wait until v1.0 and a port though.