logging missed poll values

[cpszx]

Hi,

am getting to grips with this, still not quite got the snmptrapd sorted, but have noticed a problem in the formatting, so will switch to a perl script instead and do it my way

I am not using any agents at all, i am doing ALL SNMP polling stats, and have noticed a confusing situation when the polled host does not respond.

If a poll response does not come back, no entry is put into the database.
This is fine for the occasional missed packet, as the graphs will smooth out.

But, when a host does not respond for a longer period, i think it should log at least something.

This show a 15 minute gap in polling, due to a host rebooting.

2004.Oct.10 14:53:34 331.7907
2004.Oct.10 14:52:51 1185.4722
2004.Oct.10 14:37:50 350.8108
2004.Oct.10 14:37:13 290.5278

However, the graph is only noticeable when viewed to the exact hour, because there is a long straight line between the events.
On the daily graph, it look like the host never went down!

Would it be possible to set a configurable option for what to log after so many missed polls ?

EG 5 missed polls then log a 0 into the database (numeric) or no response (character)

Or make it a default to log 0 or nothing, but still make a log entry, rather than make no log entry at all.

Regards,
Craig

[Alexei]

But, when a host does not respond for a longer period, i think it should log at least something.

Why? To get rid of straight lines?

I think this problem has to be fixed by making graphs show nothing instead of the long straight lines. Am I right?

Besides, you may always check availability of a host using a trigger similar to:

{host:status.last(0)}=2

[cpszx]

Why? To get rid of straight lines?

I think this problem has to be fixed by making graphs show nothing instead of the long straight lines. Am I right?
[/i]

Yes, thats correct. If nothing is returned, then something should go into the database so that it can be graphed correctly, rather than missing a whole chunk of data over that timescale.

The fact that the graph shows a line between the last returned value before host down, and next value returned, does not reflect that there was actually NO traffic during that time. It "looks" like there was traffic when there wasn't.

Besides, you may always check availability of a host using a trigger similar to: {host:status.last(0)}=2

historical graphed data is the easiest way to make the point to the viewer though.

Regards,
Craig