[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Probing more than once?

From: Jan-Henrik Haukeland
Subject: Re: Probing more than once?
Date: Thu, 14 Apr 2005 00:46:07 +0200

On Apr 14, 2005, at 0:09, Martin Pala wrote:

It could be possible ... it will be needed to use data payload of icmp packet to store timestamp of sending here (it will allow to analyse request/response packet pairs).

Current test is simple - we sent just one packet and expect the response, no payload is needed.

I'm +1 to add this feature to the wishlist ...

Here's mine +1. We should also consider instead of sending just one packet, we should send at least 10. It's easy to loose one package somewhere in the network.


Stefan Alfredsson wrote:
I use monit with ICMP to monitor the liveness of remote (a.k.a. some 15
hops away) machines.
I very often get false positives because of lost probes in
the middle network, compared to running monit on the same LAN.
As a workaround, I use the following:
 check host with address
   start program = "/bin/true"
   stop  program = "/bin/true"
   if 3 restarts within 4 cycles then timeout
        if failed icmp type echo with timeout 20 seconds then restart
if failed host port 80 protocol http then restart
So, three out of four probes would then trigger a timeout event.
This is however a bit kludgy, as fake restarts need to be done to get to
the timeout event. It also take three cycles before discovery.
I also get spammed with 'restart' notifications (but they can be disabled of
Is it possible to do this another way? I would imagine something like
 if three failed icmp type echo with timeout 5 seconds then ...
 where 'three failed' should indicate three consecutive attempts of
probing, and if one of the succeeds, all is well. If all fails, then
Perhaps it can be added to the wishlist? :-)
To unsubscribe:

To unsubscribe:

Jan-Henrik Haukeland
Mobil +47 97141255

reply via email to

[Prev in Thread] Current Thread [Next in Thread]