Re: {bb} bb frequency

[Philip Clark <bblist@cannae.uklinux.net>]

On Wed, 2006-08-23 at 12:53, Steve.Sayler@noaa.gov wrote:
> Thanks for your concern, The default bb time cycle, testing once every 5
> minutes, falls outside the range for my organization's status monitoring
> requirements, which recommends a 1 minute testing interval.
> 
> The re-creation of web page output is not an issue, like testing is,
> since an email notificaton would occur upon failure indication. But if
> the 1-minute testing testing overhead is detrimental to bb's
> performance, perhaps we can re-think the requirements.

In part, testing at these frequencies is likely  to be detrimental
to any monitoring option. That is because there is no inherent
difference (in most network protocols) between failing to respond
and responding very slowly. Consequently, any sufficiently robust
network test must take a relatively long time to deal with an
unresponsive host. This can be offset in BB, to some extent, by
running more network testing "threads" (at the cost of increased
load).

Stock (BTF) BB also sacrifices a fair amount of efficiency in order
to gain superior flexibility and avoid dependencies. If appropriate,
much of this can be clawed back with things like fping and bbgen.
It also helps to have (lots of) "big iron" to run it on and a
fast network.

That said, a good deal of time and expense can be avoided by
by careful planning. I often approach this by considering which
error conditions are the most likely and which would be considered
critical. In some circumstances, connectivity is king and running
fping more frequently than normal will get the job done. Alternatively,
it may be that web server response or logfile analysis are crucial
and an external script would be appropriate.

There may also be a political dimension to this. Some of the aversion
to reducing the cycle time might be because it is often requested by
"management" who can pluck numbers from the air without regard to the
consequences (until they see the cost).

It also seems strange that you mention e-mail notification. If this
is your primary method, I would give serious consideration to changing
that. Because (on a modern network) e-mail is effectively instant it
masks the fact that the underlying design is fairly relaxed about
delivery time. It is often the case that, if a mail server is heavily
loaded when a message arrives, it will be queued and retried at the
next queue cycle. The default for sendmail used to be 30 minutes later!

Cheers, Phil.



-- 
Research is what I'm doing when I don't know what I'm doing.
(Wernher von Braun)

--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-=
To unsubscribe from this list, or to subscribe to the bb-digest list
send e-mail to mailto:majordomo@bb4.com with unsubscribe bb -and/or-
subscribe bb-digest in the BODY of the message.