Small warehouse plus my house next door. Six ESP8266 nodes (temperature, two door contacts, a freezer current clamp, a gate sensor). Each one currently connects to Hermetry by itself over MQTT/TLS.
Problem: the warehouse internet drops a few times a week, sometimes for an hour. When it does, the nodes lose everything, and one of them also drives a local buzzer that I’d like to keep working (gate open + freezer warm = beep).
I found uMQTTBroker, which runs a broker on an ESP8266. Plan would be: one ESP8266 as local broker, nodes publish to it, something forwards to Hermetry when the internet is up. Has anyone done this? Is an ESP8266 enough for 6 clients? And can Hermetry’s broker bridge to it or receive retained messages after an outage?
Personally I wouldn’t put the one piece that must stay up on the weakest chip in the building. But curious what others did.
uMQTTBroker works, I ran it for a while. No TLS, and it fell over at about 8 clients for me.
what happens to the Hermetry alarms when the internet is down? do they just not fire?
@Rachel they can’t, Hermetry never sees the data until it comes back. Same reason we keep local interlocks on the shop floor and use the cloud for notices and history. Your buzzer logic belongs on site.
I tried exactly the uMQTTBroker thing at home last year and gave up on it, then did this, which has been solid for 8 months. Not an expert, so take it as “what worked for me”.
A Pi Zero 2 W running Mosquitto is the local broker. The ESP8266 nodes publish to it on the LAN, plain 1883 is fine there because it never leaves the house. The buzzer logic is a tiny Python script on the Pi that subscribes locally, so it works with no internet at all.
For Hermetry, the Pi is one device marked as a gateway (the “This device is a gateway” switch when you add it). A second script subscribes to everything on the local broker, stamps each reading with the Pi’s time (NTP-synced, check timedatectl), keeps them in a SQLite file, and sends them up with paho-mqtt over TLS:
topic: v1/gw/telemetry
{
"freezer-1": [{"ts": 1762941600000, "values": {"temp": -18.6, "amps": 1.92}}],
"gate": [{"ts": 1762941605000, "values": {"open": 0}}]
}Each node becomes a child device by its name. It only deletes a row from SQLite after the PUBACK. When the internet comes back it sends the backlog oldest first, in batches, and the readings land at their real times in history.
Two things I learned the hard way. Hermetry ignores the retain flag and doesn’t bridge to your broker, so don’t build on either. And a big backlog drains at the profile’s Replay messages per second rate, so after a long outage it takes a while to catch up, that’s normal.
The page I followed was Devices behind a gateway.
- You must be logged in to reply to this topic.
Join the conversation
Sign in to ask and answer, like helpful posts and mark the reply that solved your problem. Every contribution earns points and levels.