[rrd-users] Gathering number of entries being loaded into SQL DB per hour.

[William R. Lorenz]

Hi Micha,

On Mon, 24 Nov 2008, Micha Drodziewicz wrote:

> William R. Lorenz, dnia 2008-11-24 17:07 napisal:

>> >  I have a problem designing rrd database. The case is I want to keep 
>> >  track of number of entries being loaded into the database. My approach 
>> >  is to query for number of inventories that were loaded from one hour 
>> >  before till NOW and then store this number. But I have a problem with 
>> >  designing DS and RRA's. Any help would be greatly appreciated.

>>  Why not simply have the DS/RRA expect an entry every hour, and then do
>>  a 'SELECT COUNT(*) FROM table' with the appropriate WHERE clause and
>>  then call-out to rrdtool as you run the query each hour?  This should
>>  be no different than using rrdtool to track other variables each hour

> I am creating RRD file with this command:

> rrdtool create db.rrd --step=3600
> DS:loads:GAUGE:4000:U:U RRA:AVERAGE:0.5:1:24

> I then update it every hour with data gathered within this hour. Why I 
> do this instead of doing a full count? Well, this is Oracle DB which is 
> 600GiB in size and I do not know if it is a good idea to query the full 
> count every time (maybe there is some Oracle expert here that will tell

A full row count and an incremental hourly count are _completely_ 
different from an SQL standpoint. :-)  I imagine an hourly count with an 
hourly gauge is what you're after, rather than calculating counter 
difference between hours, as you're looking for rows added per hour?

> The problem is that I get sloppy edges. When value changes from 100 to 
> 200 I get 100, 110, 190 in the rrd data. So there is no sharp peak.

I don't understand what you mean by "sharp peak" here.  Perhaps someone 
with more rrdtool experience will chime-in with a better answer for you.

-- 
William R. Lorenz