[smokeping-users] Re: ظ´£º Re: Forks & monitoring data reliability

[Niko Tyni]

On Tue, Jun 07, 2005 at 12:08:09PM +0800, Joe Shen wrote:

> Yes, I read DNS.pm code and found it probe target
> server in a for() loop but not sending 20 queries in
> parallel. 

The 20 pings for one target are always sent sequentially, and the
result is then the average of these. As I said, that's a feature.  :)

> > > 2. How to set up timeout value to reach most
> > reliable
> > > probing data set?
> > 
> > It should be the number of seconds without response
> > after which you
> > decide that the server is down and is not going to
> > respond at all.
> > 10 might be a bit high here, but YMMV.
> 
> What value do you use for your DNS server checking? 10
> seconds means the total time allowed for 20 probes or
> for each probe?

It's the timeout for one probe (or 'ping' or 'query'). This means that
in the worst case, for pings = 20, it can take 200 seconds to complete
the probing of an unresponsive server. 

Now, if forks < targets, some targets are going to be probed
sequentially, and if there are two unresponsive targets in a row that
will make one round of probing take about 400 seconds, which is over
the default step value (300 seconds). If this happens, the next round
is skipped for all targets (because the previous measurement is still
running.)

I'm using the default values myself, but I'm only probing a couple
of targets. 

See the 'basefork.pm' documentation (and code :) for more information.

Cheers,
-- 
Niko

--
Unsubscribe mailto:smokeping-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:smokeping-users-request at list.ee.ethz.ch?subject=help
Archive     http://lists.ee.ethz.ch/smokeping-users
WebAdmin    http://lists.ee.ethz.ch/lsg2.cgi