Would you use a self-hosted geofencing solution?

[toggledbits]

Some of you may know that I took at shot at building an alternate geofencing solution for Vera. The core of it was system agnostic, using the OwnTracks application and AWS lambdas to track devices and keep a central data, then disseminate that to the Vera via a websocket-based plugin. It worked with other apps as well, including Tasker and GPSLogger, but of the dozen people that were testing it, most used OwnTracks.

A lot was learned in the process, not the least of which is that the success of any such solution is highly dependent on the phone and its settings. Phone manufacturers love to set things up for the longest battery life, of course, but that's usually very anti-geofencing behavior. In the case of at least one brand, it was unusable and the settings could not be modified. It was also cost-prohibitive to maintain on Amazon, as AWS grabs a dime here and a dollar there and before you know it, it added $100/month to my AWS bill, which my wife deducted from my Scotch budget. Unacceptable.

But it's quite reasonable to use OwnTracks to a local endpoint, and I could pretty easily replicate the functionality as a local application, or maybe even as an additional endpoint built into MSR's API (still separate port and process, but in the package).

So the question really is... would you do it, or would you be too concerned about the security risks associated (e.g., dynamic DNS and NAT mapping in the firewall necessary for the phone to contact the service when not on LAN)?

[rafale77]

Scotch budget...lol

I definitely would have been all in if I had not had already a better alternative. The NAT mapping would be my biggest concern if i was to do something like this and I would have used a VPN instead which may or may not be practical but maybe safer and not especially harder to setup. (probably easier actually).

Since I also caught an allergy to Google a couple of years ago due to some of their insidious behaviors, I don't have any android device and likely never will so @amg0's iphoneLocator works wonderfully well. I have gone as far as to fork the plugin to replace google map on it with 2 alternatives... The main benefits of that approach is the absence of any need for additional mobile app and minimal additional cloud service (still need apple and a mapping service) along with stunning configurability (I am using it to get a TTS announcement to remind me to recharge my phone when I am at home and my battery runs low and use the distance to trigger various modes) so I am probably not your best candidate but anything locally hosted gets my vote.

[Pabla]

I second what @rafale77 has said. I use native Homekit geofencing to toggle a VSwitch. Its bulletproof and anyone can get it working with a Homekit bridge. As for Androids I have only used geofencing once on a Galaxy S10, and its been pretty reliable there too.

[RogerO]

As you know from our previous email conversation, I would definitely use a local hosted version. I hate Apple more than Google so my wife and I have Samsung phones. They worked well with Logo when you had it running.

[Matteburk]

For me I use Hubitat (ex vera) with homebridge and let Native Apple geofence do the trick . It has been the best solution I have tested and super easy to operate. I dont think I will try another Solution if not the apple one will break.
/Mattias

[therealdb]

I’m using owntracks pushing to an Azure Function (like aws lambdas) that’s pushing messages into a queue. I’m using it in general for cloud to home messages. I’m transforming these messages into mqtt messages from my app monitoring the same queue and then executing actions against endpoints from the inside. It’s instant and it’s costing me less than 1 usd per month, since messages queue is very cheap. It’s a bit overkill, since it’s an architetture we usually design for very high troughput, but it’s very simple and effective. It’s pub sub but in the cloud.

[iRobot]

At the moment I'm one of the lucky few where Vera geofencing works most of the time. In fact, that's the only function I'm using actively with my plus.

I most certainly would by interested in a local geofencing solution. I probably will be using a VPN or maybe setup a separate internal network or vlan with NAT for it.

Was planning on trying tasker or owntracks, but that's still in the idea fase in my head.

[LibraSun]

Ha, I just concluded several weeks' thinking on HOW to remove the Geotags from my Vera Plus which had been introduced by the VeraMobile app. Mission accomplished: Go to app > Menu > Geofencing > Enable (requires app Location "Always" permissions) > see formerly hidden geotags > Delete > disable > remove permissions > done!

To answer your question above, YES, I'd be interested. I might also let loose some pocket change to help make it happen.

And I know I'm in the small minority of users who really don't mind Google or whoever else tracking my whereabouts; I leave all tracking turned on both in mobile and Web settings for most products I use (not random sites!), and love Maps / MyTracks / Strava / etc.

[CatmanV2]

I'd be interested. I already run DynDNS for mail services and a VPN endpoint, which I assume would obviate the need for the NAT port mapping (although I do have SSH open, just in case)

C

[gwp1]

I just stumbled across this thread and will proudly own up to being one of those dozen or so who beta'd Lobo and had great success with it. I never had false triggers using that. FFWD to today and I'm doing a mashup of multiple sources from Homekit to OwnTracks to GeoFency to (experimenting with) an iBeacon - the goal being "don't open my garage door when I'm outside the geofence in the next neighborhood over pedaling my bike around."

All these words, I guess, can be summed up in "yes, I'd be interesting."

[tamorgen]

I'm just getting started using MSR. I'm currently using iPhoneLocator as a plugin on Vera, but I'm looking eventually to retire my Vera and move to Ezlo. Ezlo has on their roadmap native Geofencing, however, if it's anything like Vera's, I don't see it being particuarlly reliable. I found iPhoneLocator to be much more reliable. Unfortunately, the developer has said he has no plans of migrating it to Ezlo, so I'm stuck with waiting for Ezlo to devlop theirs, or keeping Vera running at the same time as Ezlo, and using iPhoneLocator as a Geofencing solution through MSR.

I would like to see something self hosted through MSR.

[CatmanV2]

iPhone locator on openLuup is fine, with the caveat that it is far far far too late to the party nearly all the time. No matter what settings I've tried to give it for update times and so on, I was home for 10 minutes on Wednesday before my phone acknowledged.

C

[therealdb]

I'm on owntracks since I've moved everything except Zwave from my Vera, and I'm OK. It's instant, so I got notifications from our system at the very same moment we're crossing boundaries. I have defined a couple of zones (Home, Office, etc) to turn on/off scenes based on the position of both me and my wife.

I'm still waiting for a 100% reliable way to detect the cars/motorbikes, but that's another story.

[CatmanV2]

Ahh I see, an active solution so I assume (without having read any of the docs) you set the fences on the device and that somehow the server communicates to the Vera

C

[tamorgen]

@therealdb said in Would you use a self-hosted geofencing solution?:

owntracks

How does it integrate with MSR or other platforms?

[CatmanV2]

Exactly what I'm trying to work out It's gonna be in here, somewhere

OwnTracks Booklet

C

[toggledbits]

Just to chime in here, since I've had some relevant experience in this area (together with some of you who went on that journey with me)...

At the time I made a geofencing solution to replace Vera's, I also used Owntracks. It had great promise, but it had some issues:

1. The iOS version had a small, enthusiastic group of devs that was active and attentive. The Android version, at the time, was supported by a single developer who wasn't very active, and refused to implement some of the newer features of the iOS version. The iOS version apparently worked really well — that was my own experience in testing with an old phone donated by one of the testers, and most iOS users didn't need a lot of support. I'm an Android guy, so I had ample Droids to test with, and it presented problems, with the app itself being the least of them. Principally, Android phone manufacturers have increasingly tweaked the phones to maximize battery life, with some of that optimization at the expense of performance. They also started to limit services for privacy requirements in various countries (iOS, too). Location services was a big area to hit, as was background processing (and location services running in the background was especially problematic), so out of the box, Android users were faced with learning where and how to modify settings deep in the bowels of the system to allow Owntracks Android to keep up with changes in both foreground and background. Some manufacturers didn't even present these options, notably some Huawei models, so Owntracks was basically unusable for the purpose on those phones. Because each Android phone manufacturer customizes Android for each model of phone, there's not only no consistency across manufacturers, but even across models within the same manufacturer, making it a heavy lift to instruct users on what settings to change and where, and troubleshoot performance and accuracy issues. The privacy issues and

2. At the time, and I don't know if this has changed (perhaps @therealdb can comment), it was necessary for Owntracks to have an endpoint on the web that it could reach to send location updates. The only way that Owntracks communicated was by MQTT or HTTP POST. That means one of two things: (1) you have to open a port in your home firewall/network for your Vera/MSR/whatever hub (or MQTT broker) to receive the update posts from the phone(s), or (2) you had to have a public HTTP(S) endpoint or MQTT broker to collect the data and relay it to the Vera/MSR/hub over some other connection. Anybody can do #1 with sufficient knowledge of networking, specifically their own router configuration, etc. In my system, I went to option #2, building a web site together with a number of lambdas in AWS that could receive the updates from the phone in a high-availability, high-performance environment, and allow the hub to communicate with that infrastructure to update the Vera plugin (for example). I called this project Lobo, and it included the site, the endpoints, and a Vera plugin. The problem here was that the AWS infrastructure ended up costing about $100/month (they literally nickel and dime you to death), and as I started to explore scaling it beyond the users that participated in the experiment, it started to look like something that I would have had to charge US$2-$3 per month per phone to maintain. Given that Android was so troublesome, I decided that there would probably be little uptake for a paid solution, so I shut the project down. I still have all the code, of course. But basically, if you don't open up your own network to allow Owntracks to get to it, then you have to have public infrastructure that can relay the data for Owntracks, and that likely means somebody has to pay for it (whether it's public infrastructure you build and maintain yourself, or a shared/cloud infrastructure that someone like me maintains and likely has to charge for).

I'm sure you can do it today with MSR and MQTTController, but you'd need to work out either the firewall access to your in-house MQTT broker, or a public MQTT broker that both MQTTController and Owntracks could reach. You just have to have sufficient knowledge on your own to do it. And if you're on an Android device, you have to figure out the settings to get it to work properly and with sufficient accuracy that false positives and negatives don't make you crazy (you have my sympathies).

[RogerO]

Anyone using Zerotier to access their internal MQTT broker without poking holes in their firewall? I am a big fan of Zerotier but haven't tried that yet.

[therealdb]

@toggledbits yeah, we’re iPhone users (as many using iPhone Locator), so we’re covered. I went for #2, with an azure function sending the payload to a queue that’s then imported into the mqtt broker. It’s probably safer to bridge a public mqtt server to your internal one or to use something to just bride one endpoint to your internal network.

I agree it’s not trivial, but home assistant should offer webhooks for this integration quite easily.