On Friday 12 August 2005 07:20 pm, Christopher Taylor wrote:
> On 8/12/05, Ivo Perich <[EMAIL PROTECTED]> wrote:
> > I have a Linux server doing ping to about 130 remote hosts. It is
> > monitoring
> > the time response and making mrtg graphics on each host. Everything works
> > ok, but there is a bureaucratic problem: in the project's documentation
> > they
> > say that the pings has to be done FROM the remote host TO the server.
> > Why? I
> > don't know, because every monitoring system i have seen does the pings
> > probes
> > or whatever itself. What do you think? There's any technical difference
> > between doing the ping from one side or another in order to monitor time
> > response? (Besides of the costs, of course. We should put a dedicated
> > machine
> > on every point.) I write this email to be used as an expert opinion in
> > front
> > of my counterpart.
> >
> > Thanks for your response
> >
> > Ivo Perich
> > BITs Ltda, Temuco, Chile
> >
> > -------------------------------------------------------
>
> Is it possible that the ping coming from the client is being used as a keep
> alive signal that allows the client to know that the server is still up?
> Also, having the pings on the clients saves CPU all around. Each client
> does one ping each instead of the server pinging each individual computer.
> This is just a thought as I have no idea as to the scope or purpose of the
> project you are working on.
Well, the fact is that our monitoring server has no problem doing the pings,
because we are doing about 15 ping every 5 minutes to each machine. The
server is ok, the network is ok. The point is that we think the monitoring
system is ok and our counterpart says that the pings must be done from the
remote hosts, where there are no machines to dedicate to do this job. We want
to do the monitoring the way we are doing it now, because we only need to
monitor the response time and the percent of succeed pings, and this can be
done from our server without need to acquire about 130 PCs that would be
doing pings to our server. The other solution is so much complicated, and we
think that is no reason to do the monitoring that way. If there's a reason,
we would like to know it. That's all. What do you think?
Thanks for your response
Ivo
The important thing here is control. It is infinitely easier to control the pings and diagnose problems from on machine. Having the pings on the clients can cause problems diagnosing problems. Also, the server needs to know if a machine is up in order to write data to it. The way to insure the machine is up is the ping. It is also much easier to gather and analyze the ping statistics if the are on one dedicated machine as opposed to distributed over the network.
--
Christopher Taylor - Registered Linux User #383327
