Re: monitoring servers

[Henrique de Moraes Holschuh <hmh@debian.org>]

On Thu, 15 Dec 2005, Walhalla wrote:
> ok - but i think we missunderstood - i need something to check 
> remote services not services on local machines. services not in that 

Hmm, why not use Net SNMP then? It lets you plug basically *anything* you
need locally to inspect stuff, and publish it over SNMP.  And *all* good
remote monitoring softwares can do SNMP.

> way if a proc is running - it is enought if the corresponding port is 
> open/running -> server ist up & running.

That is not good enough a test in my experience, but if that's all you need,
even socat can be used to create a quick script to check for open ports.
And nagios has a TCP connect test you can use very easily, too.

I know from experience that if you are monitoring a web server, you NEED to
send it a http request and check for the reply if you want to know whether
the service is really up and how responsive it is.  For email, you need a
full roundrip check (send - receive), etc.   A TCP connect won't tell you if
the application behind the open port is working (or overloaded and thus not
providing new services, etc).

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh