Home Monitoring Re-Write Number Four!

Since getting the Tesla Powerwall installed, our trusted Wattson has not been able to display correct figures as it can’t tell if we are importing or exporting until the Powerwall is full.  The Wattson displays a relatively static value of +150W indicating that we’re importing, yet the data from the various other devices in the house contradicts that figure.

So it’s time to say goodbye to Wattson and hand it on to a neighbour and hope they get some use out of it.

Wattson’s demise is a great excuse to upgrade to a tablet and display a lot more information than just whether we’re importing or exporting, so I’ve gone out and bought a Samsung Galaxy Tab A from JL to replace Wattson.

In order to display more information on the tablet, I needed to re-write the home monitoring application and start graphing the data at home rather than relying on PVOutput.  PVOutput is a great website, but it’s limited to a 5 minute picture of what’s going on and I’ve run out of fields to upload data, even though I donate to get extra fields! Wattson has gotten us used to being able to see what’s going on instantly rather than waiting for a snapshot 5 minutes later.

The second re-write I did of the home monitoring application in 2015 has been running well for a few years, but despite what I wrote back then about it being maintainable, it was a pain to add in a new datasource and it was written in my least favourite framework – Mule.

Since then I’ve tried re-writing it in Node.js, but that code was less than elegant and not tested at all… It also relied on a heavy weight MySQL database which I wanted to avoid if possible. HSQLDB may be a bit basic, but it’s served me well for many years and allows me to make changes to the files in a text editor if required.

I did learn something valuable from the Node.js re-write – consolidate the five tables I had before into one large table. I’ve changed the following five tables

to a single table for ease of storing the data and to save space.

The previous database file size was 640MB (note that’s more than 200MB per year as I blogged about the database being 400MB only last year) vs. the new single table layout file size of 240MB. Every field in the database except the composite primary keys are nullable. This allows the data to be stored into the table in any order, after all I can’t guarantee which Arduino will send it’s data first.

The next step was to work out how to convert the database from the original layout to the new layout without having my pc running at 100% for over 2 hours (the first time I loaded the data from the old tables to the new table, this is exactly what happened!). The trick was to not insert based on a select union, but to use the HSQLDB merge functionality. The two hour ETL turned into a three minute ETL. This much improved ETL time allows me to take a copy of the old database (the in use one) at any time, transform it and check the new app is compatible with the schema and can write data into the new layout correctly.

As I’ve mentioned above, the new application is no longer based on Mule and instead is a Spring Boot app.   The home monitoring application receives input using Spring MVC controllers and persists the data to the database against the date and time (rounded to the minute).

At the service layer, there’s also three separate scheduled services, one for uploading PVOutput data once a minute, one for requesting the EE addons status page and scraping the data every hour and one for calling the Tesla Powerwall API every five seconds.

EE addons status page scraping I hear you say… “what’s that for?”  We no longer have fixed line internet and rely on EE 4G internet, which is great until we run out of data two days before the end of the month!  The EE addons status page displays how much data you have used, how much is remaining and how long until the next period.  Since I’ve now got the option to display a lot of different data on the tablet, it seemed sensible to display the EE data allowance too!

For anyone interested in doing something similar, here’s a class I’ve written to read the HTML and trim it to extract the right bits of information. The fields aren’t accessible as I don’t store the information – I simply pass it straight to Splunk via toString.

package uk.co.vsf.home.monitoring.service.ee;

import java.util.regex.Matcher;
import java.util.regex.Pattern;

import org.apache.commons.lang3.StringUtils;
import org.apache.commons.lang3.builder.ReflectionToStringBuilder;
import static org.apache.commons.lang3.StringUtils.*;

public class EeDataStatus {

	private static final String ALLOWANCE_LEFT = "allowance__left";
	private static final String ALLOWANCE_TIMESPAN = "allowance__timespan";
	private static final String BOLD_END = "</b>";
	private static final String BOLD_START = "<b>";
	private static final String SPAN_END = "</span>";
	private static final String SPAN_START = "<span>";
	private static final String DOUBLE_SPACE = "  ";

	private final String allowance;
	private final String remaining;
	private final String timeRemaining;

	public EeDataStatus(final String response) {
		String allowance = response.substring(response.indexOf(ALLOWANCE_LEFT) + ALLOWANCE_LEFT.length());
		allowance = allowance.substring(0, allowance.indexOf(SPAN_END));

		Pattern pattern = Pattern.compile("(\\d+.*\\d*GB)");
		Matcher matcher = pattern.matcher(allowance);

		matcher.find();
		this.remaining = matcher.group();
		matcher.find();
		this.allowance = matcher.group();

		String timespan = response.substring(response.indexOf(ALLOWANCE_TIMESPAN) + ALLOWANCE_TIMESPAN.length());
		timespan = timespan.substring(0, timespan.indexOf(SPAN_END));
		timespan = timespan.substring(timespan.indexOf(SPAN_START) + SPAN_START.length());
		timespan = timespan.replaceAll(BOLD_END, EMPTY).replaceAll(BOLD_START, EMPTY);
		timespan = timespan.replaceAll(CR, EMPTY);
		timespan = timespan.replaceAll(LF, EMPTY);
		timespan = timespan.replaceAll(DOUBLE_SPACE, SPACE);
		timespan = StringUtils.trim(timespan);
		this.timeRemaining = timespan;
	}

	@Override
	public String toString() {
		return new ReflectionToStringBuilder(this).toString();
	}
}

When I tried writing the home monitoring application in Node.js I gave Prometheus a go to see whether that would be a good tool for graphing at home.  It worked well when graphing small sets of data, but when I tried to graph over a years worth of data, it either errored because there was too much data coming back from the query, or took a vast amount of time to refresh the graph.  It’s possible I wasn’t using the tool correctly, but I decided it wasn’t for me in this use case because of the inability to graph large amounts of data and because it’s not as intuitive as the graphing tool I’ve chosen to go with.

So what graphing tool have I chosen?  Splunk 🙂

I chose Splunk for a number of reasons:

  1. I’ll be sending less than 500MB to Splunk a day, so it’s free 😀
  2. It’s incredibly intuitive to search through data in Splunk, so I should be able to give my dad a basic lesson and he can create graphs for himself. I had considered the ELK stack, but the searching language isn’t quite as intuitive…
  3. Splunk doesn’t care about the schema of the data you throw at it.  This makes it easy to work with as I can add/remove fields when required and not have to change a schema.

Writing the data to Splunk uses the ToStringBuilder JSON format and a Log4j socket appender.  The ToStringBuilder format is configured at bootup via the following component.

package uk.co.vsf.home.monitoring;

import org.apache.commons.lang3.builder.ToStringBuilder;
import org.apache.commons.lang3.builder.ToStringStyle;
import org.springframework.stereotype.Component;

@Component
public class ToStringBuilderStyleComponent {

	public ToStringBuilderStyleComponent() {
		ToStringBuilder.setDefaultStyle(ToStringStyle.JSON_STYLE);
	}
}

And I chose the Log4j socket appender because it doesn’t require the use of tokens to talk to Splunk.

<?xml version="1.0" encoding="UTF-8"?>
<Configuration status="warn">
    <Appenders>
        <Socket name="socket" host="SERVER NAME" port="9500">
            <PatternLayout pattern="%m%n"/>
        </Socket>
        <Console name="STDOUT" target="SYSTEM_OUT">
        </Console>
    </Appenders>
    <Loggers>
        <Logger name="uk.co.vsf.home.monitoring" level="info" additivity="false">
            <AppenderRef ref="socket" />
            <AppenderRef ref="STDOUT" />
        </Logger>

	...

</Configuration>

Bringing it all together, we’ve gone from Wattson which displayed only one figure – house load – as shown in the (albeit not great) picture below:

To this 😀

And this complicated device/application diagram

Hopefully this incarnation of the home monitoring application will last a few years, but I suspect I’ll be re-writing it all again at some point 🙂

References
Tesla Powerwall 2 API https://github.com/vloschiavo/powerwall2/
Log4j2 Socket Appender https://logging.apache.org/log4j/2.x/manual/appenders.html#SocketAppender

Hot Water Tank Temperature on Arduino LCD

Back in March 2014, I posted about monitoring the hot water tank temperature and when the immersion was switched on by the solar surplus http://blog.v-s-f.co.uk/2014/03/home-monitoring-home-made-results/.

We have now decided to display the last temperature reading on an LCD next to the shower room and boiler control panel. The previous software I’d written to upload data to PVOutput had* to be re-written in order to accommodate this new Arduino which would request the last reading available and display it.

The Arduino code below is fairly simple. Every 1 minute it requests data from my server named Pompeii. If it receives data, it refreshes the screen with the response, changing the last character for an up or down arrow, to represent the temperature in/decreasing or a hyphen for level. It’s possibly not the most efficient code in the world, but it doesn’t need to be!

There is one notable item in the code and that’s the pin configuration of the LCD. It turns out that the pins referenced in the Arduino tutorial book I have are pins used for the Ethernet connection on the Ethernet board, therefore I’ve shifted pins 11 and 12 to 6 and 7 respectively.

So what does it look like?

IMG_5312

Code below

#include <SPI.h>
#include <Ethernet.h>
#include <LiquidCrystal.h>

///////// CHANGEABLE VALUES /////////

char pompeii[] = "192.168.0.16";
int pompeiiPort = 28080;

double minutesBetweenCalls = 1.0;

///////// CHANGEABLE VALUES ABOVE /////////

EthernetClient pompeiiClient;
byte mac[] = {
  0x90, 0xA2, 0xDA, 0x0F, 0xA1, 0xCF};
char pompeiiService[] = "/aggregator/services/hot-water-display";
double millisecondsInAMinute = 60000.0;

//LiquidCrystal lcd(12, 11, 5, 4, 3, 2);
LiquidCrystal lcd(7, 6, 5, 4, 3, 2);

const int LCD_WIDTH=16;

byte upArrow[8] = {
  B00100,
  B01110,
  B11111,
  B10101,
  B00100,
  B00100,
  B00000,
};

byte downArrow[8] = {
  B00000,
  B00100,
  B00100,
  B10101,
  B11111,
  B01110,
  B00100,
};

void setup() {
  lcd.createChar(1, upArrow);
  lcd.createChar(0, downArrow);
  lcd.begin(16, 2);
  lcd.setCursor(0,0);

  Serial.begin(9600);

  delay(1000);

  connectToEthernet();
}

void connectToEthernet() {
  // attempt to connect to Wifi network:
  // start the Ethernet connection:
  if (Ethernet.begin(mac) == 0) {
    Serial.println("Failed to configure Ethernet using DHCP waiting 1 minute");
    delay(millisecondsInAMinute);

    if (Ethernet.begin(mac) == 0)
    {
      Serial.println("Failed to configure Ethernet using DHCP waiting 1 more minute");
      delay(millisecondsInAMinute);

      if (Ethernet.begin(mac) == 0) {
        Serial.println("Failed to configure Ethernet using DHCP stopping - will need reset");
        while(true);
      }
    }

  }
  // give the Ethernet shield a second to initialize:
  delay(1000);
  Serial.println("connecting...");
  //lcd.print("connecting...");

  Serial.print("Connected to the network IP: ");
  Serial.println(Ethernet.localIP());
  //lcd.print("Network up...");
}

void loop() {
  displayData();

  Serial.println("fetched data");
  delay(millisecondsInAMinute * minutesBetweenCalls);
}

void displayData() {
  String data = receiveDataFromPompeii();

  if(data.length() > 0) {
    lcd.clear();
    lcd.home();

    String lastChar = "";
    if(data.endsWith(" U") || data.endsWith(" D") || data.endsWith(" L")){
      lastChar = data.substring(data.length() - 1);
      data = data.substring(0, data.length() - 1);
      Serial.println("last char " + lastChar);
    }
    Serial.println("display data " + data);

    if(data.length() > LCD_WIDTH) {
      lcd.home();
      lcd.print(data.substring(0, LCD_WIDTH));
      lcd.setCursor(0, 1);
      if(data.length() > (LCD_WIDTH * 2)) {
        lcd.print(data.substring(LCD_WIDTH, LCD_WIDTH * 2));
      }
      else {
        lcd.print(data.substring(LCD_WIDTH));
      }
    } 
    else {
      lcd.setCursor(0, 0);
      lcd.print(data);
    }

    if(lastChar == "U") {
      lcd.print((char) 1);
    } 
    else if (lastChar == "D") {
      lcd.print((char) 0);
    }
    else if (lastChar == "L") {
      lcd.print("-");
    }
  }
}

String receiveDataFromPompeii() {
  Serial.println("Request data from Pompeii");

  String response = "";

  if (pompeiiClient.connect(pompeii, pompeiiPort)) {
    Serial.println("connected to pompeii");
    // Make a HTTP request:
    pompeiiClient.print("GET ");
    pompeiiClient.print(pompeiiService);
    pompeiiClient.println(" HTTP/1.1");
    pompeiiClient.print("Host: ");
    pompeiiClient.print(pompeii);
    pompeiiClient.print(":");
    pompeiiClient.println(pompeiiPort);
    pompeiiClient.println("Accept: text/html, text/plain");
    pompeiiClient.println("Pragma: no-cache");
    pompeiiClient.println("Cache-Control: no-cache");
    pompeiiClient.println("Connection: close");
    pompeiiClient.println();

    Serial.println("Called pompeii");
    delay(3000);

    String dataRead = "";
    boolean reachedData = false;
    while (pompeiiClient.connected() || pompeiiClient.available()) {
      char c = pompeiiClient.read();
      //Serial.print(c);

      if( reachedData ) {
        response += c;
        //Serial.println(response);
      }

      dataRead += c;

      if(dataRead.endsWith("\r\n\r\n")) {
        Serial.println("Reached the data");
        reachedData = true;
      }
    }
    Serial.println("Finished reading data");

    pompeiiClient.stop();
    pompeiiClient.flush();
    Serial.println("Closed connection");
  }

  Serial.println("response data to return is " + response);
  return response;
}

* As any programmer knows, when you look at your old code, you think… “what on earth!” But also there was a technical reason for re-writing part of it (although I ended up re-writing it all) and that was to store the readings more locally rather than having to go out onto the internet to get the readings.