[rrd-users] RRD Graphs with bad averaging of missing data?

[Alex van den Bogaerdt]

> just after an outage,
> the next update is mostly in the next bucket but a tiny fraction is inside
> the bucket that contains mostly the time where an outage occured.  As a
> result that bucket has a very low value instead of 'unknown'.

and, as Steve correctly said, this will also happen if you give zeros to 
RRDtool. Worse: my trick won't work in that case.

Can you show us the data you have in your plain text file, a few lines will 
do, starting two updates just before such an outage at your 9:56-ish and 
ending two updates just after the same outage.