[rrd-developers] Re: Posible bug/unexpected behaviour on templated update

[javier at rediris.es]

> > Last week, I tried to modify rrdupdate to met our needs and I discovered
> > the problem. In whichever way you call rrdupdate, all datasources are
> > set to unknown, and later the command line values are feeded, rewritting
> > the previous unknown values. As every datasource last_ds is rewritten,
> > those not included in the template gets an unknonw last_ds, leading
> > to uknown counter/derive values.
> 
> if you do not include a variable in the template its value is
> assumed to be UNKNOWN. This is the intended behaviour ... updates
> to single rrd databases are atomic in the sense that all values
> must be updated together ...

We use a custom udp client/server to collect the values, and the size
of the message in some cases was too long, and the easier way is to
call twice to rrdupdate with different datasources in any case. One of
my ideas before reading the code and realizing about the unkowns, was
to make a change to allow updating with the same second that last update
and add a flag to delay the consolidation (both pdp and cdp).
Reading the source code, that looks a extreme complication, and allowing
a partial update showed to me as much easier and much powerful. It looks
like that the only (and also strong) point where the full-update is assumed
is the use of last_up to calculate the time intervals.

In my opinion it is not hard to avoid this full-update, but it could be a
little harder (but probably possible) to avoid it without any change in the
actual rrd structures. The point is if should be valuable the extra effort
of the second path or the increased flexibility is not valuable or even
desirable.


								Javi



--
Unsubscribe mailto:rrd-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-developers-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-developers
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi