Hacker News Reader

> and nothing depends on some company's servers.

Exactly my motivation. I have an mqtt broker at a heart of everything. homekit2mqtt for homekit/siri integration, lirc2mqtt so my harmony remote can join in, node-red providing some automation logic, etc.

But where it gets interesting is being able to support devices that won't otherwise. Homekit support for Philips Hue means €50 for a new hub - or mqtt-hue so I can support it myself. Belkin Wemo doesn't support homekit, and has no plans to - but I've done it via node-red regardless.

DIY looks messy from the outside. But once the ball is rolling, anywhere a vendor won't support the features I require, I'll support them myself. The ability to do so makes it all worth it.

> DIY looks messy from the outside. But once the ball is rolling, anywhere a vendor won't support the features I require, I'll support them myself. The ability to do so makes it all worth it.

Pretty sure the messy part isn't about writing code, it's about maintaining it. What happens if one of the APIs you're talking to changes? What happens if you want to add a new device into your ecosystem? If there's a power outage and one of your hubs doesn't come back up, how much time/energy does it take to figure out which one and to bring it back up?

Not saying that it's not a worthwhile project, just that it takes a lot of effort. That said, I have my own growing list of bots that helps me automate repetitive tasks in my life that I maintain...

If an API changes, I can catch up with it in a quiet afternoon - instead of buying whatever hardware they tell me their next update requires.

Catching dark hardware is really easy - just allow each component to die if it can't reach the hardware, and then catch their 'last will' messages in mqtt (mqttwarn is great for this, I can get an email whenever a 'last will' is published). Also a problem the vendors won't solve for me ..

I roll my own home automation with a combination of things. Once it is up I don't know that it I'm any worse off than anyone else. The worst I had was the weather API I was using dried up. 20 minutes later I had weather from somewhere else. I like that if I end up with a pile of useless junk it is my fault, not the decision of company that decided to stop supporting their hardware and shut down their cloud servers [1]. I don't need the cloud and I don't need a 3rd part to have control of my garage door opener. I'm still waiting for the hack that opens a half million garage doors around the country! (I have no idea how many doors these big companies control, but it would make great news.)

This may be the biggest reason I go DIY on my automation - it gives me a chance to learn things that I wouldn't get into at work.

[1] https://www.theguardian.com/technology/2016/apr/05/revolv-de...

I too replaced my dying PoS Rainbird irrigation controller with a raspberry pi and a relay board.

> With cron, I can schedule things however I want!

Wait, are you not using sprinklers_pi[1]? I highly recommend it.

I have more home pi/arduino/IoT projects coming up, but I am thinking it's going to be a lot of work integrating them. I'd like one home app/page that integrates them all.

1. https://github.com/rszimm/sprinklers_pi/wiki

What did you dislike about your Rainbird system ?

1. It didn't work reliably

2. The UI was awful and buggy

3. It finally broke outright and replacements with halfway modern features were pricey

I installed my own sprinklers and drip - 22 zones of it. A Rainbird controller was going to be nearly $300. I went with a pi and loved it. Then came the real problem - I put the home on the market and ended up replacing my pi with a rainbird for the next people. Why? I kinda wanted to keep my setup and buyers didn't know what to think of it.

After I swapped it out I missed quite a few things - controlling it from my phone, from home or remote, being able to pull up a graph and see how long each zone ran. I wasn't in the mood to get up at 3am to see if the rainbird was working so I crossed my fingers and hoped it would work at least until the house sold. Setting up all my zones was a lot faster on the pi than it was with cranky dials and buttons on the rainbird. I won't miss it.

For the integration side you should take a look at http://home-assistant.io - they support a huge variety of hardware already, and almost every type of device includes a configurable MQTT adapter if you're building your own.

Once your hardware is integrated with Home Assistant it can handle automating things. The whole system is a big state machine so most automations will involve reacting to state changes and sending commands to other devices in response.

If it is only for irrigation of anything not grass, why not use off the grid system like blumats (http://blumat.com/en/)? You only need to provide it with a water source.

Having a pi cluster "feels" more like a datacenter with many servers. Using kubernetes to coordinate it all is pretty fun when you actually have several "servers" to coordinate across. That to me was the satisfying part. Not that I NEEDED miniature scale, but that the miniature scale felt like a valid downsizing of real scale.

I've been tempted to do something like this, what are you doing about shared storage?

I have a nas on the network as well. That isn't set up quite like it would be in a real datacenter, but it works for centralizing data

>I still need to work out how to copy the images up to S3 or a Linode. I'm using MotionEyeOS and it doesn't seem to know how to do that itself //

Can you just make the folder it copies to a link to a mounted remote filesystem, mount your server folder over SSH, for example.

Controlling the sprinkler with the Pi is a cool idea!

> The second is a security camera for a vacation rental home, and is not quite done yet.

What is the purpose of the camera?

I run a script that takes a picture with an old Kinnect and pushes to dropbox. It is a dirty solution but it worked...until I ran out of space...