RRDtool data inconsistency - rrdtool

Question

I am using RRDtool to graph the status of a pump on my Raspberry Pi. I must be doing some configuration wrong, since the values are close to what I am inputting, but not exact.
The pin status should be either 1 or 0.
<!-- 2014-03-10 10:24:00 CDT / 1394465040 --> <row><v>NaN</v></row>
<!-- 2014-03-10 10:25:00 CDT / 1394465100 --> <row><v>NaN</v></row>
<!-- 2014-03-10 10:26:00 CDT / 1394465160 --> <row><v>1.0000000000e+00</v></row>
<!-- 2014-03-10 10:27:00 CDT / 1394465220 --> <row><v>2.3711630000e-01</v></row>
<!-- 2014-03-10 10:28:00 CDT / 1394465280 --> <row><v>9.8168226667e-01</v></row>
<!-- 2014-03-10 10:29:00 CDT / 1394465340 --> <row><v>1.6624716667e-02</v></row>
<!-- 2014-03-10 10:30:00 CDT / 1394465400 --> <row><v>9.8544061667e-01</v></row>
<!-- 2014-03-10 10:31:00 CDT / 1394465460 --> <row><v>2.9590616667e-02</v></row>
<!-- 2014-03-10 10:32:00 CDT / 1394465520 --> <row><v>9.7204963333e-01</v></row>
<!-- 2014-03-10 10:33:00 CDT / 1394465580 --> <row><v>2.6263616667e-02</v></row>
<!-- 2014-03-10 10:34:00 CDT / 1394465640 --> <row><v>9.7533411667e-01</v></row>
<!-- 2014-03-10 10:35:00 CDT / 1394465700 --> <row><v>2.3075633333e-02</v></row>
<!-- 2014-03-10 10:36:00 CDT / 1394465760 --> <row><v>9.7849575000e-01</v></row>
<!-- 2014-03-10 10:37:00 CDT / 1394465820 --> <row><v>1.9948233333e-02</v></row>
<!-- 2014-03-10 10:38:00 CDT / 1394465880 --> <row><v>9.8158333333e-01</v></row>
<!-- 2014-03-10 10:39:00 CDT / 1394465940 --> <row><v>1.6888216667e-02</v></row>
<!-- 2014-03-10 10:40:00 CDT / 1394466000 --> <row><v>9.2141166667e-01</v></row>
<!-- 2014-03-10 10:41:00 CDT / 1394466060 --> <row><v>5.2411610000e-01</v></row>
<!-- 2014-03-10 10:42:00 CDT / 1394466120 --> <row><v>5.2411610000e-01</v></row>
<!-- 2014-03-10 10:43:00 CDT / 1394466180 --> <row><v>9.6672030000e-01</v></row>
<!-- 2014-03-10 10:44:00 CDT / 1394466240 --> <row><v>5.0939110833e-01</v></row>
<!-- 2014-03-10 10:45:00 CDT / 1394466300 --> <row><v>5.0939110833e-01</v></row>
<!-- 2014-03-10 10:46:00 CDT / 1394466360 --> <row><v>4.9845539167e-01</v></row>
<!-- 2014-03-10 10:47:00 CDT / 1394466420 --> <row><v>4.9845539167e-01</v></row>
<!-- 2014-03-10 10:48:00 CDT / 1394466480 --> <row><v>9.9399037500e-01</v></row>
<!-- 2014-03-10 10:49:00 CDT / 1394466540 --> <row><v>9.9399037500e-01</v></row>
<!-- 2014-03-10 10:50:00 CDT / 1394466600 --> <row><v>2.6977033333e-02</v></row>
<!-- 2014-03-10 10:51:00 CDT / 1394466660 --> <row><v>9.7898348333e-01</v></row>
<!-- 2014-03-10 10:52:00 CDT / 1394466720 --> <row><v>9.7898348333e-01</v></row>
create_db.sh
#!/bin/bash
rrdtool create pinstats.rrd \
--step 60 \
DS:pump:GAUGE:600:0:1 \
RRA:MAX:0.5:1:2016
update.sh
#!/bin/sh
a=0
while [ "$a" == 0 ]; do
echo "pump ondate"
/home/pi/on.sh
/home/pi/graph.sh
pump=1
rrdtool update pinstats.rrd N:$pump
sleep 60
echo "pump offdate"
/home/pi/off.sh
/home/pi/graph.sh
pump=0
rrdtool update pinstats.rrd N:$pump
sleep 120
done

Answer 1

You are being affected by Data Normalisation.
This will adjust your values according to a linear approximation in order to make the time point lie on an interval boundary. IE, if your Interval is 5min, then the updated value will be as it was at 12:00, 12:05, 12:10 ... etc
This makes sense if you are graphing a large number which is a rate; the overall averages still work out and the data are at regular intervals. However, if you are using a Gauge data type with small integers it becomes problematic.
In order to avoid this, you have to update on an interval boundary rather than using N as your time point.
Try this shell code:
interval=60
timestamp=`date +%s`
num_intervals=`expr $timestamp / $interval`
adjusted_time=`expr $num_intervals '*' $interval`
rrdtool update pistats.rrd $adjusted_time:1
sleep $interval
adjusted_time=`expr $adjusted_time + $interval`
rrdtool update pistats.rrd $adjusted_time:0
This code ensures that the update times are exactly on an interval boundary, and hence no Data Normalisation is performed (actually, it becomes a null operation).
For more details, see Alex van den Bogaerdt's excellent tutorial here.