This program enables the ESP8266 to become the central node in a small distributed IoT system. It implements an MQTT Broker and a simple scripted rule engine with event/action statements that links together the MQTT sensors and actors. It can act as STA, as AP, or as both and it can connect to another MQTT broker (i.e. in the cloud). Here it can act as bridge and forward and rewrite topics in both directions. Also it can parse JSON structures, send basic HTTP GET requests and do basic I/O: i.e. read and write to local GPIO pins, react on timers and GPIO interrupts, drive GPIO pins with PWM, and read the ADC.
In the user directory there is the main program that serves as a stand-alone MQTT broker, client and bridge. The program starts with the following default configuration:
This means it connects to the internet via AP ssid,password and offers an open AP with ap_ssid MyAP. This default can be changed in the file user_config.h. The default can be overwritten and persistenly saved to flash by using a console interface. This console is available either via the serial port at 115200 baud or via tcp port 7777 (e.g. "telnet 192.168.4.1 7777" from a connected STA).
- set [ssid|password] _value_: changes the settings for the uplink AP (WiFi config of your home-rou- set [ap_ssid|ap_password] _value_: changes the settings for the soft-AP of the ESP (for your stations)ter)
- set ap_on [0|1]: selects, whether the soft-AP is disabled (ap_on=0) or enabled (ap_on=1, default)
- set ap_open [0|1]: selects, whether the soft-AP uses WPA2 security (ap_open=0, automatic, if an ap_password is set) or open (ap_open=1)
- set auto_connect [0|1]: selects, whether the WiFi client should automatically retry to connect to the uplink AP (default: on=1)
While the user interface looks similar to my esp_wifi_repeater at https://github.com/martin-ger/esp_wifi_repeater this does NO NAT routing. AP and STA network are stricly separated and there is no routing in between. The only possible connection via both networks is the uMQTT broker that listens on both interfaces.
The broker comes with a "local" and a "remote" client, which means, the broker itself can publish and subscribe topics. The "local" client is a client to the own broker (without the need of an additional TCP connection).
By default the "remote" MQTT client is disabled. It can be enabled by setting the config parameter "mqtt_host" to a hostname different from "none". To configure the "remote" MQTT client you can set the following parameters:
The esp_uMQTT_broker comes with a build-in scripting engine. A script enables the ESP not just to act as a passive broker but to react on events (publications and timing events), to send out its own items and handle local I/O. Details on syntax and semantics of the scripting language can be found here: https://github.com/martin-ger/esp_mqtt/blob/master/SCRIPTING.md . Examples of scripts are in the "scripts" directory.
For this, you first have to make the scripts available via a web server. A quite simple way to do that is to start a minimal web server in the "scripts" directory, i.g. on Linux you can start a server on port 8080 on the delevopment machine with:
The ESP tries to download the script from the given URL and prints upon success or failure a report on the console. Currently the download only works via plain HTTP and no redirects are followed.
Now the ESP listens on the given port for an incoming connection and stores anything it receives as new script. Upload a file using netcat, e.g.:
```bash
$ netcat 192.168.178.29 2000 <user/demo_script2
```
The ESP will store the file and immediatly checks the syntax of the script:
```
CMD>script 2000
Waiting for script upload on port 2000
Script upload completed (451 Bytes)
Syntax okay
CMD>
```
You can examine the currently loaded script using the "show script" command. It only displays about 1KB of a script. If you need to see more, use "show script <line_no>" with a higher starting line. Newly loaded scripts are stored persistently in flash and will be executed after next reset if they contain no syntax errors. "script delete" stops script execution and deleted a script from flash.
NTP time is supported and timestamps are only available if the sync with an NTP server is done. By default the NTP client is enabled and set to "1.pool.ntp.org". It can be changed by setting the config parameter "ntp_server" to a hostname or an IP address. An ntp_server of "none" will disable the NTP client. Also you can set the "ntp_timezone" to an offset from GMT in hours. The system time will be synced with the NTP server every "ntp_interval" seconds. Here it uses NOT the full NTP calculation and clock drift compensation. Instead it will just set the local time to the latest received time.
After NTP sync has been completed successfully once, the local time will be published every second under the topic "$SYS/broker/time" in the format "hh:mm:ss". You can also query the NTP time using the "time" command from the commandline.
The code can be used in any project that is compiled using the NONOS_SDK or the esp-open-sdk. Also the complete broker in the user directory can be build using the standard SDKs after adapting the variables in the Makefile.
Thanks to Tuan PM for sharing his MQTT client library https://github.com/tuanpmt/esp_mqtt as a basis with us. The modified code still contains the complete client functionality from the original esp_mqtt lib, but it has been extended by the basic broker service.
The broker does support:
- MQTT protocoll versions v3.1 and v3.1.1 simultaniously
- a smaller number of clients (at least 8 have been tested, memory is the issue)
The MQTT server will now run in the background and you can connect with any MQTT client. Your Arduino project might do other application logic in its loop.
The complete broker functionality is included in the mqtt directory and can be integrated into any NONOS SDK (or ESP Arduino) program ("make -f Makefile.orig lib" will build the mqtt code as a C library). You can find a minimal demo in the directory "user_basic". Rename it to "user", adapt "user_config.h", and do the "make" to build a small demo that just starts an MQTT broker without any additional logic.
in the "user_init()" (or Arduino "setup()") function. Now it is ready for MQTT connections on all activated interfaces (STA and/or AP). Please note, that the lib uses two tasks (with prio 1 and 2) for client and broker. Thus, only task with prio 0 is left for a user application.
With these functions you can publish and subscribe topics as a local client like you would with any remote MQTT broker. The provided dataCb is called on each reception of a matching topic, no matter whether it was published from a remote client or the "MQTT_local_publish()" function.
If an *MqttAuthCallback* function is registered with MQTT_server_onAuth(), it is called on each connect request. Based on username, password, and optionally the connection info (e.g. the IP address) the function has to return *true* for authenticated or *false* for rejected. If a request provides no username and/or password these parameter strings are empty. If no *MqttAuthCallback* function is set, each request will be admitted.
The *MqttConnectCallback* function does a similar check for the connection, but it is called right after the connect request before any internal status is allocated. This is done in order to reject requests from unautorized clients in an early stage.
To adjust memory consumption of one MQTT connection and thus the max number of concurrent connections you can redefine MQTT_BUF_SIZE and QUEUE_BUFFER_SIZE in "user_config.h". MQTT_BUF_SIZE is the max. size of pending inbound messages for one connection (and thus also the max. size of a single MQTT message) and QUEUE_BUFFER_SIZE is the max. size of all pending outbound messages for one connection. Currently these parameters are set to 1024 resp. 2048 bytes, resulting in a memory consumption of about 4 KB per connection and a max number of connections of about 8-9 (depending on the memory usage of the rest of the program). When you reduce buffer sizes, e.g. to 512 and 1024 bytes, a single connection requires only about 2.5 KB resulting in up to 13 possible concurrent MQTT connections. In any case you have to increase the number of TCP connections (default 5) first by calling "espconn_tcp_set_max_con(n)" with n, the max. number of concurrent TCP connections, less or equal to 15.
Also there is a hard limitation on the number of STAs connected to the SoftAP, which is 8. I.e. when using the esp_uMQTT_broker only with clients via the SoftAP interface, even with reduced memory consumtion, the limit of different client nodes is still 8, as it is imposed by the binary WiFi driver. Only when used via the STA interface and an external AP you can connect more than 8 MQTT clients.
