Automating my TV

One of the lazy things that I've tried to do was have the Amazon Echo turn my TV on and off. When I had Home Assistant running on my Raspberry Pi, I used a component that controlled the TV and Apple TV via HDMI CEC. Unfortunately it wasn't quite reliable and I lost the ability to use it when I migrated to a VM for Home Assistant.

In a recent release of Home Assistant, support was added for Roku and since I have a TCL Roku TV, I decided to give it a try. The component itself works, but has a few major limitations for me. First off it initializes on Home Assistant startup. In order to conserve a little energy, I have my TV, Apple TV, and sound bar on a Z-Wave controlled outlet. The outlet doesn't turn on until the afternoon, so most of the time when Home Assistant restarts (I have it restart at 6 am so that my audio distribution units initialize as they also turn off at night), the TV isn't turned on. The second issue has to do with the TV going to sleep. It has a deep sleep and a fast start mode; fast start uses more energy, so I leave it off. The Roku component uses HTTP commands to control the device or TV; when the TV is in deep sleep, it doesn't respond to HTTP commands. This, of course, makes it impossible to turn on the TV with the component.

After thinking about this problem for awhile, I came up with some Node-RED flows to turn on the TV and handle status updates. The TV, it turns out, responds to a Wake-On-LAN packet as I have it connected via Ethernet and Home Assistant has a WOL component that lets me send the packet.

My flow to check on the TV state is a bit complicated.

1. First it pings the TV. The ping is done every 10 seconds.
2. If the TV responds, it sends an HTTP request to the TV.
3. When the response comes back, it is parsed, the current application running is checked. This also lets me know what Roku channel is currently active. I have noticed that my TV reports that the Davinci Channel is active when I turn the TV off, so I special case that.
4. If the channel is not null and not the Davinci Channel, I then send a command to check to see if the display is off.
5. After I figure out the app and if the display is off, I craft a new payload with the current channel in it.
6. The payload is then sent in an HTTP request back to Home Assistant's HTTP Sensor API
7. If the TV doesn't respond to the ping, I set the payload to off and then send the state to the Home Assistant API.

Turning on the TV is a bit less complicated.

1. Send WOL packet to TV.
2. Pause.
3. Send HTTP command to turn on TV.
4. Send HTTP command to set input to HDMI3 (my Apple TV).

Turning off the TV is even easier.

1. Send HTTP command to turn off TV.

When I turn on the TV outlet, the state of the TV gets updated pretty quickly as the ping command from above is running every 10 seconds.

I've posted the Node-RED flows below that can be imported and modified for your situation.

Download Node-RED flow to turn Roku TV on/off

Download Node-RED flow to get current TV state

Adding Energy Monitoring to Home Assistant

Now that I have Home Assistant running pretty well, I've started seeing what else I can add to it. There are several components for monitoring energy usage, but sadly none for my Rain Forest Automation Eagle Energy device. After a quick search, I found a Node-RED flow that looked promising. It would query the local API of the device (or the Cloud one) and give me an answer. The next step was seeing how to get that into Home Assistant. I found the HTTP Sensor which would let me dynamically create sensors. (There is so much to explore in Home Assistant, it will keep me entertained for awhile.)

With all the pieces in place, I set the Node-RED flow to repeat every 15 seconds and then use the Home Assistant API to add the usage. This worked well for a few hours, but then I stopped receiving updates only to discover that the Eagle device had stopped responding. When I checked the developer documentation, it indicated that the local API was not supported for my older device. However, an Uploader API was available that would push the data to me. That sounded interesting, so I created a flow in Node-RED that listened for a connection and then parsed the data using the flow I found before. The only problem was how do I get the device to push me the data. The Eagle device has options for cloud providers, but something I missed before is that there is the ability to add a new one. I added my Node-RED install using: http://10.0.3.100:1880/eagleenergy and the data starting flowing!

So far this new method has been working for almost 24 hours, so I have a lot more confidence that this will keep working.

The Node-RED flow is below:

    [{"id":"b7dc8932.f85388","type":"xml","z":"15825822.ae8e1","name":"Convert to
    Object","property":"payload","attr":"","chr":"","x":310,"y":1440,"wires":[["21269998.3f97e6"]]},{"id
    ":"c85cb9e0.0ba818","type":"http request","z":"15825822.ae8e1","name":"Query Active
    App","method":"GET","ret":"txt","url":"http://10.0.3.41:8060/query/active-app","tls":"","x":310,"y":
    1380,"wires":[["b7dc8932.f85388"]]},{"id":"b35dc59a.083738","type":"ping","z":"15825822.ae8e1","name
    ":"Ping
    TV","host":"10.0.3.41","timer":"10","x":90,"y":1300,"wires":[["e2decb33.dc36e8"]]},{"id":"e2decb33.
    dc36e8","type":"switch","z":"15825822.ae8e1","name":"Is TV
    responding?","property":"payload","propertyType":"msg","rules":[{"t":"false"},{"t":"else"}],"
    checkall":"true","repair":false,"outputs":2,"x":290,"y":1300,"wires":[["6b2354bd.8306ec"],["c85cb9e0
    .0ba818"]],"outputLabels":["","Is awake?"]},{"id":"a59cc790.158f68","type":"http
    request","z":"15825822.ae8e1","name":"Update TV
    State","method":"POST","ret":"txt","url":"https://homeassistant.com:8123/api/states/sensor.tv_state"
    ,"tls":"","x":1140,"y":1440,"wires":[[]]},{"id":"208bd98e.9ed806","type":"change","z":"15825822.
    ae8e1","name":"Set API
    Authorization","rules":[{"t":"set","p":"headers","pt":"msg","to":"{\"Authorization\":\"Bearer
    INSERT_BEARER\"}","tot":"json"}],"action":"","property":"","from":"","to":"","reg":false,"x":920,"y"
    :1440,"wires":[["a59cc790.158f68"]]},{"id":"21269998.3f97e6","type":"change","z":"15825822.ae8e1","
    name":"Set Payload to Active
    App","rules":[{"t":"set","p":"payload","pt":"msg","to":"payload.active-app.app[0]._","tot":"msg"}],"
    action":"","property":"","from":"","to":"","reg":false,"x":330,"y":1500,"wires":[["6d6fdcd0.004714"]
    ]},{"id":"71eea783.5158f8","type":"function","z":"15825822.ae8e1","name":"Set new
    payload","func":"var channel = msg.payload\nif (channel == \"Davinci Channel\") {\n    channel =
    \"Idle\"\n}\nvar newPayload = {\"state\":channel,\"attributes\":{\"icon\":\"mdi:television\",
    \"friendly_name\":\"TV\"}}\nmsg.payload = newPayload\n\nreturn
    msg;","outputs":1,"noerr":0,"x":650,"y":1400,"wires":[["208bd98e.9ed806"]]},{"id":"6b2354bd.8306ec",
    "type":"change","z":"15825822.ae8e1","name":"Set Channel to
    off","rules":[{"t":"set","p":"payload","pt":"msg","to":"off","tot":"str"}],"action":"","property":""
    ,"from":"","to":"","reg":false,"x":870,"y":1300,"wires":[["71eea783.5158f8"]]},{"id":"ee5484a0.
    bf69c8","type":"http request","z":"15825822.ae8e1","name":"Get Device
    Info","method":"GET","ret":"txt","url":"http://10.0.3.41:8060/query/device-info","tls":"","x":300,"y
    ":1580,"wires":[["dfba9256.d0175"]]},{"id":"6d6fdcd0.004714","type":"switch","z":"15825822.ae8e1","
    name":"Check active app","property":"payload","propertyType":"msg","rules":[{"t":"eq","v":"Davinci
    Channel","vt":"str"},{"t":"null"},{"t":"else"}],"checkall":"true","repair":false,"outputs":3,"x":610
    ,"y":1500,"wires":[["ee5484a0.bf69c8"],["ee5484a0.bf69c8"],["71eea783.5158f8"]],"outputLabels":["
    Davinci
    Channel","Null",""]},{"id":"dfba9256.d0175","type":"xml","z":"15825822.ae8e1","name":"Convert to
    Object","property":"payload","attr":"","chr":"","x":510,"y":1580,"wires":[["37ef8802.728648"]]},{"id
    ":"37ef8802.728648","type":"change","z":"15825822.ae8e1","name":"Check Display
    State","rules":[{"t":"set","p":"payload","pt":"msg","to":"payload.device-info.power-mode[0]","tot":"
    msg"}],"action":"","property":"","from":"","to":"","reg":false,"x":560,"y":1660,"wires":[["adaad7e0.
    489ba8"]]},{"id":"adaad7e0.489ba8","type":"switch","z":"15825822.ae8e1","name":"Is Display
    Off?","property":"payload","propertyType":"msg","rules":[{"t":"eq","v":"DisplayOff","vt":"str"}],"
    checkall":"true","repair":false,"outputs":1,"x":760,"y":1660,"wires":[["6b2354bd.8306ec"]]}]

All In with Home Assistant

I've spent parts of the last 9 months playing with Home Assistant and have written about some of my adventures. A few weeks ago, I finally decided to go all in with Home Assistant and ditch my Vera. I bought an Aeotec Z-Stick Gen5 Z-Wave dongle and starting moving all my devices over to it. Within a few days, I had all my devices moved over and unplugged my Vera. Everything was running great on my Raspberry Pi B, but I noticed that the History and Logbook features were slow. I like looking at the history to look at temperature fluctuations in the house.

I had read that switching from the SQLite database to a MySQL database would speed things up. So I installed MariaDB (a fork of MySQL) on my Raspberry Pi and saw a slight increase in speed, but not much. Next was to move MariaDB to a separate server using Docker. Again, a slight increase in speed, but it still lagged. At this point everything I read pointed to running Home Assistant on an Intel NUC or another computer. I didn't want to invest that kind of money in this, so I took a look at what I had and started down the path of installing Ubuntu on my old Mac mini which was completely overkill for it (Intel Quad Core i7, 16 GB RAM, 1 TB SSD). Then I remembered that I had read about a virtual machine image for Hass.io and decided to give that a try.

After some experimenting, I managed to get Home Assistant installed on a virtual machine running in VMWare on my Mac Pro. (A few days after I did this, I saw that someone posted an article documenting this.) I gave the VM 8 GB of RAM, 2 cores (the Mac Pro has 12) and 50 GB of storage. Wow, the speed improvement was significant and history now shows up almost instantly (the database is running in a separate VM)! I was so pleased with this, I decided to unplug the Raspberry Pi and make the virtual machine my home automation hub. There were a few tricks, however. The virtual machine's main disk had to be setup as a SATA drive (the default SCSI wouldn't boot), suspending the VM confused it, and the Z-Wave stick wouldn't reconnect upon restart. After much digging, I found the changes I needed to make to the .vmx file in the virtual machine:

    suspend.disabled = "TRUE"
    usb.autoConnect.device0 = "name:Sigma\ Designs\ Modem"

(The USB auto connect is documented deep down on VMWare's site.)

I've rebooted the Mac Pro a few times and everything comes up without a problem very quickly, so I'm now good to go with this setup. Z-Wave takes about 2.5 minutes to finish startup vs. 5 or 6 on the Pi. A friend asked if I was OK with running a "mission critical" component on a VM. I said that I was because the Mac Pro has been rock solid for a long time and my virtual machines have been performing well. I could change my mind later on, but I see no reason to spin up another machine when I have a perfectly overpowered machine that is idle 95% of the time.

What next? Now that I have more power for my automation, I may look at more pretty graphs and statistics. I may also just cool it for awhile as I've poured a lot of time into this lately to get things working to my satisfaction. This has definitely been an adventure and am glad that I embarked on it.

Dipping my toe in the world of Docker

A former co-worker of mine has talked about Docker for years and I've taken a look at it a few times, but have generally been uninterested in it. Recently with my interest in Home Assistant, I've decided to take another look as many of the installs of Home Assistant as well as Hass.io are based on Docker.

I've used virtual machines running on VMware Fusion for years with some Windows installs and some Linux installs. I'm very comfortable with Linux, but kind of dislike maintaining different packages. There are package managers that handle much of it for me, but then there are other packages that have special installations.

I had a few goals in mind for seeing if Docker could replace the current virtual machines I had running for Pi-hole and Observium. The goals were pretty simple that I wanted easy updates and be able to easily backup the data. In the Docker world, updates are dead simple and in many docker containers, the data is stored outside of the container making it easy to backup. As another goal, I wanted to be able to experiment with other containers to see what else I could add to my network.

With all this in mind, I started looking at how to setup Docker. Pretty quickly, I realized that Docker for the Mac was virtually useless for me as it didn't handle all the networking that Docker running on Linux could. So that meant installing Docker on a Linux VM; that almost negated my goal of easy updates as I'd still have to update the virtual machine running Ubuntu. I could live with that if the rest of the setup was straight forward and didn't have to remember how to update each container individually.

In order to make backups easy, I wanted to store the data on my Mac and not inside of the virtual machine. I've not had great luck with the VMWare tools for mounting volumes, so I decided to use CIFS (SMB) to mount a volume in Linux which works well except for the MariaDB (MySQL fork) Docker container. Not a big deal, I'd just add a cron job to dump the databases every few hours and store the dumps on the mounted volume. I added the following to /etc/fstab

    //myserver/account/Documents/Ubuntu /mnt/mediacenter cifs username=account,domain=WORKGROUP,password=password,rw,hard,uid=1000,gid=1000 0 0

I also had to turn on Windows File Sharing options on the Mac.

The crontab is:

    30 */2 * * * /usr/local/bin/backup_mysql

with the backup_mysql file being

    #!/bin/sh
    /usr/bin/mysqldump -h 127.0.0.1 -u root -ppassword --lock-all-tables --all-databases | gzip > /mnt/mediacenter/backups/mysql/mysql_backup_$(date +"%m-%d-%Y-%H_%M_%S").gz
    find /mnt/mediacenter/backups/mysql/* -mtime +3 -exec rm {} \;

The next hurdle was dealing with IPv6; most people don't care about it, but I'm not most people! IPv6 is quite complicated (at least to me), so that took a bit of experimenting to get it to work in Docker. For future reference, ndppd lets the virtual machine tell the world that it handles IPv6 for the Docker containers (basically).

So where was I? After getting the Linux VM setup, it was on to setting up my containers. With docker-compose, I could setup one file that was the configuration for all my containers. Now this was great as I could modify it and test out different containers. After a few days of work, this is the core of my docker-compose file. There are a few other containers I've added including LibreNMS, but this is basically what I have. The nginx-proxy is great as I just add DNS entries for each service and it handles SSL and lets me run multiple web services on the same machine.

version: "2.3"
services:
  nginx-proxy:
   image: jwilder/nginx-proxy
   environment:
      - DEFAULT_HOST=pihole.exmple.com
   ports:
     - "80:80"
     - "443:443"
     - "::1:8080:80"
   dns:
     - 10.0.1.1
   volumes:
     - /var/run/docker.sock:/tmp/docker.sock:ro
     - '/mnt/mediacenter/docker/certs:/etc/nginx/certs'
   restart: always
   networks:
      default:
        ipv6_address: XXXX:XXXX:XXXX:XXXX:1::2