[rrd-users] RRD Not Storing Historical Gauge Data

Alex van den Bogaerdt alex at vandenbogaerdt.nl
Wed Nov 5 19:14:42 CET 2008

----- Original Message ----- 

>> But it appears all of your updates are more than 120 seconds apart.
> That was it.  I had thought the script that performs the iterative tests
> was running more frequently.  With the updated HB, it works like a champ.


> For a gauge-based application such as this, whereas the times gauge

that does not matter. gauge, counter and so on decide how rates are 
computed, nothing more.

> readings are available may vary between 1-45 minutes, would there be any
> harm in setting a high heartbeat, such as 3600 seconds (60 minutes)?  I

No, not at all. As long as you understand what's happening, you can set it 
to anything appropriate for your situation.  If you need to have unknown 
data (e.g. at reboot time) you can always force an unknown by updating with 
value "U".

> would assume the average would be calculated between steps/intervals, per
> the documentation?

It would compute one rate (using gauge you're just telling it the rate) and 
then normalize that rate into many intervals (timespans defined by 'step'). 
That could mean you would be getting { 1000, 1000, 1000, 1000 } instead of 
{ 1200, 900, 900, 1000 } if you would have updated four times instead of one 

> Would a heartbeat value of '0' cause RRD not to apply
> heartbeat NaNs, and as such the average would be taken for all intervals?

I don't remember zero being a special value, but to be 100% sure you would 
need to examine the source. You could always set it to 604800 (one week) or 
even more.

More information about the rrd-users mailing list