Re: keepalive in socket server program

[Scott Klement <rpg400-l@xxxxxxxxxxxxxxxx>]

It depends on what is meant by "keepalive"?

See here for a description of keepalive at the TCP layer vs higher layers:
https://en.wikipedia.org/wiki/Keepalive

If they mean keepalive at the TCP layer, that's easy enough.  You can 
enable that with the setsockopt() API and the TCP_KEEPALIVE option.  (I 
would recommend setting everything explicitly in your program rather 
than relying on the CHGTCPA setting.)    The rest is done for you under 
the covers.

If they want keepalives at the application-level (which are, arguably, 
more useful) then you need to build that into your communications 
protocol. Since they're giving you specifications, I would guess that 
they will tell you how the protocol (and also keep alives) work in that 
case.

Since you're writing a server program, please consider using INETD 
rather than writing the whole thing yourself.  (No need to muck about 
with spawn, sendmsg, or take/givedescriptor that way...)




On 8/28/2015 3:35 PM, Jeff Young wrote:
> I am working on a socket server program where the connection will remain
> open most of the day.  The specifications from the client (I have no
> control over this) indicate that they will periodically send a keepalive to
> me.
> My questions are: 1 - How do I detect that a keepalive was sent
>                             2 - What (if anything) do I need to send back.
>
>
> TIA
>
>
> Jeff Young
> Sr. Programmer Analyst