Receiving Events
A few things to consider at the receiving side when forwarding events with Data Connectors.
Both the header and body of the incoming request contain information of interest that can be extracted and, depending on the configuration, used to verify the content and origin of the request.
- X-Dt-Signature Includes a JSON Web Token (JWT). Once decoded using the signature secret, it contains all the necessary information to verify the request content and origin.
Note that depending on the framework used to receive the Data Connector events, the header name casing may differ. Some services will force header name to lower-case, like
x-dt-signature.The request body contains three fields,
event, labels, and metadata. The following snippet shows an example request body of a touch event for a humidity sensor forwarded by a Data Connector.{
"event": {
"eventId": "<EVENT_ID>",
"targetName": "projects/<PROJECT_ID>/devices/<DEVICE_ID>",
"eventType": "touch",
"data": {
"touch": {
"updateTime": "2021-05-28T08:34:06.225872Z"
}
},
"timestamp": "2021-05-28T08:34:06.225872Z"
},
"labels": {
"room-number": "99"
},
"metadata": {
"deviceId": "<DEVICE_ID>",
"projectId": "<PROJECT_ID>",
"deviceType": "humidity",
"productNumber": "102081"
}
}
Field | Type | Description |
event | struct | Contains event data. See the Event documentation where the structure is explained in detail for each event type. |
labels | struct | Device label key- and value pairs included by the Data Connector. See the Advanced Configuration page for details about include labels. |
metadata | struct | Contains metadata about the event. See the section below for more details. |
Each event has a
metadata field that includes details about the event. The event metadata has the following structure.Field | Type | Description |
|---|---|---|
deviceId | string | The identifier of the device that published the event. |
projectId | string | The identifier of the project the device is in. |
deviceType | string | |
productNumber | string |
Note
The structure of the
metadata field might change in the future if event types are added that are not published by devices. Make sure to first check event.eventType to make sure it is a known device event before processing the metadata field.See the code sample below for an example of how to do this.
The event metadata makes it possible to check which device type has published the event, even for event types like
touch or networkStatus which are published by many types of devices. This makes it possible to add new devices to a database without having to first look up the device using the REST API.The metadata also provides a more convenient way to get the
deviceId and projectId of the device that published the event, without having to parse the event.targetName field.A request-reply flow on the Endpoint URL should be implemented as follows:
- 1.Your endpoint receives an HTTPS POST request.
- 2.Your service processes the data in some way.
- 3.Your service replies to the event request with a 200 OK response.
What is important to note here is that the request should never return an HTTP 200 OK response before you are done processing it. When the DT Cloud receives a response with a 200 status code, the event will be taken off the internal Data Connector queue and checked off as received.
Best Practice
Do not reply 200 OK until you have finished processing your data.
Note that any status code in the range 2xx will be accepted as OK in the response.
When using a Signature Secret, the X-Dt-Signature header is included and contains a JWT, signed by the Signature Secret. Inside, a checksum of the request body can be found and used to check for tampering.
The following steps sum up the process of verifying the received request at the receiving endpoint.
- 1.Extract the signed JWT from the HTTP header X-Dt-Signature of the received request.
- 2.Verify the JWT's signature with the signature secret.
- 3.Calculate a SHA256 checksum over the entire request body.
- 4.Compare the body checksum with the
checksum_sha256field contained in the JWT (there's achecksumfield as well that uses SHA1 which is less secure than SHA256, and is kept only for backward compatibility). - 5.If these checksums are identical, you can be certain that the event has not been tampered with and originated from your Data Connector.
The following snippet from our Google Cloud Function example integration implements this verification process.
Python 3.9
Python API
Node.js 16
Go 1.16
# This Python is built on Flask, docs are available here:
# https://flask.palletsprojects.com/en/2.0.x/quickstart/
import os
import hashlib
import jwt # pip install pyjwt==2.3.0
from flask import Flask, request # pip install Flask==2.0.2
app = Flask(__name__)
# Read environment variable.
SIGNATURE_SECRET = os.environ.get('DT_SIGNATURE_SECRET')
@app.route('/', methods=["POST"])
def data_connector_endpoint():
# Extract the body as a bytestring and the signed JWT.
# We'll use these values to verify the request.
payload = request.get_data()
token = request.headers['x-dt-signature']
# Verify request origin and content integrity.
if not verify_request(payload, token):
return ('Could not verify request.', 400)
# We now know the request came from DT Cloud, and the integrity
# of the body has been verify. We can now handle the event safely.
handle_event(request.get_json())
# Respond with a 200 status code to ack the event. Any states codes
# that are outside the 2xx range will nack the event, meaning it will
# be retried later.
return ('OK', 200)
def verify_request(body, token):
"""
Verifies that the request originated from DT, and that the body
hasn't been modified since it was sent. This is done by verifying
that the checksum field of the JWT token matches the checksum of the
request body, and that the JWT is signed with the signature secret.
"""
# Decode the JWT, and verify that it was signed using the
# signature secret. Also verifies that the algorithm used was HS256.
try:
payload = jwt.decode(token, SIGNATURE_SECRET, algorithms=["HS256"])
except Exception as err:
print(err)
return False
# Verify the request body checksum.
m = hashlib.sha256()
m.update(body)
checksum = m.digest().hex()
if payload["checksum_sha256"] != checksum:
print('Checksum Mismatch')
return False
return True
def handle_event(body):
"""
Processes the event itself. For this example, we will just
decode a touch event, and print out the timestamp, device ID,
and the device type.
"""
# First, check if the event type is one of the event
# types we're expecting.
# As an example, we'll check for touch events here.
if body['event']['eventType'] == 'touch':
# Now that we know this is a device event, we can
# check for the device type and device identifier
# in the event metadata.
device_type = body['metadata']['deviceType']
device_id = body['metadata']['deviceId']
timestamp = body['event']['data']['touch']['updateTime']
print(f"Received touch event at {timestamp} from {device_type} sensor with id {device_id}")
# This Python is built on Flask, docs are available here:
# https://flask.palletsprojects.com/en/2.0.x/quickstart/
import os
from dtintegrations import data_connector, provider
from flask import Flask, request # pip install Flask==2.0.2
app = Flask(__name__)
# Read environment variable.
DT_SIGNATURE_SECRET = os.getenv('DT_SIGNATURE_SECRET')
@app.route('/', methods=['POST'])
def dataconnector_endpoint():
# Validate and decode the incoming request.
payload = data_connector.HttpPush.from_provider(
request,
provider=provider.FLASK,
secret=DT_SIGNATURE_SECRET,
)
# Print the event data.
handle_event(payload)
# If all is well, return 200 response.
return ('OK', 200)
def handle_event(payload):
"""
Processes the event itself. For this example, we will just
decode a touch event, and print out the timestamp, device ID,
and the device type.
"""
# First, check if the event type is one of the event
# types we're expecting.
# As an example, we'll check for touch events here.
if payload.event.event_type == 'touch':
# Now that we know this is a device event, we can
# check for the device type and device identifier
# in the event metadata.
metadata = payload.get_device_metadata()
if metadata != None:
device_type = metadata.device_type
device_id = metadata.device_id
timestamp = payload.event.data.timestamp
print(f"Received touch event at {timestamp} from {device_type} sensor with id {device_id}")
const crypto = require('crypto')
const express = require('express') // npm install [email protected]
const jwt = require('jsonwebtoken') // npm install [email protected]
// Read environment variable
const signatureSecret = process.env.DT_SIGNATURE_SECRET
// dataConnectorEndpoint receives, validates, and returns a response
// for the forwarded event.
const dataConnectorEndpoint = (req, res) => {
// Validate request origin and content integrity.
let token = req.headers['x-dt-signature']
if (verifyRequest(JSON.stringify(req.body), token) === false) {
res.sendStatus(400)
return
}
// We now know the request came from DT Cloud, and the integrity
// of the body has been verify. We can now handle the event safely.
handleEvent(req.body)
// Respond with a 200 status code to ack the event. Any states codes
// that are outside the 2xx range will nack the event, meaning it will
// be retried later.
res.sendStatus(200);
};
// Verifies that the request originated from DT, and that the body
// hasn't been modified since it was sent. This is done by verifying
// that the checksum field of the JWT token matches the checksum of the
// request body, and that the JWT is signed with the signature secret.
const verifyRequest = (payload, token) => {
// Decode the JWT, and verify that it was signed using the
// signature secret. Also verifies that the algorithm used was HS256.
let decoded
try {
decoded = jwt.verify(token, signatureSecret, { algorithms: "HS256" })
} catch(err) {
console.log(err)
return false
}
// Verify the request body checksum.
const hash = crypto.createHash("sha256")
const checksum = hash.update(payload).digest("hex")
if (checksum !== decoded.checksum_sha256) {
console.log('Checksum Mismatch')
return false
}
return true
}
// handleEvent processes the event itself. For this example,
// we will just decode a touch event, and print out the timestamp,
// device ID, and the device type.
const handleEvent = (payload) => {
// First, check if the event type is one of the event
// types we're expecting.
// As an example, we'll check for touch events here.
switch (payload.event.eventType) {
case 'touch':
// Now that we know this is a device event, we can
// check for the device type and device identifier
// in the event metadata.
const deviceType = payload.metadata.deviceType
const deviceId = payload.metadata.deviceId
const timestamp = payload.event.data.touch.updateTime
console.log(`Received touch event at ${timestamp} from ${deviceType} sensor with id ${deviceId}`)
break
default:
break
}
}
// Sets up a bare-bones server that listens on port 8080, and routes
// all requests to the path "/" to the `dataConnectorEndpoint` function
const app = express()
app.use(express.json())
app.listen(8080)
app.post('/', dataConnectorEndpoint)
package main
import (
"crypto/sha256"
"encoding/hex"
"encoding/json"
"fmt"
"io/ioutil"
"net/http"
"os"
// go get github.com/golang-jwt/jwt/[email protected]
jwt "github.com/golang-jwt/jwt/v4"
)
// Read environment variable
var signatureSecret = os.Getenv("DT_SIGNATURE_SECRET")
// DataConnectorEndpoint receives, validates, and returns a response
// for the forwarded event.
func DataConnectorEndpoint(w http.ResponseWriter, r *http.Request) {
// Extract the body and the signed JWT.
// We'll use these values to verify the request.
tokenString := r.Header.Get("x-dt-signature")
bodyBytes, err := ioutil.ReadAll(r.Body)
if err != nil {
fmt.Println(err)
http.Error(w, err.Error(), http.StatusBadRequest)
return
}
// Validate request origin and content integrity.
if err := verifyRequest(bodyBytes, tokenString); err != nil {
fmt.Println(err)
http.Error(w, err.Error(), http.StatusBadRequest)
return
}
// We now know the request came from DT Cloud, and the integrity
// of the body has been verify. We can now handle the event safely.
if err := handleEvent(bodyBytes); err != nil {
fmt.Println(err)
http.Error(w, err.Error(), http.StatusBadRequest)
return
}
// Respond with a 200 status code to ack the event. Any states codes
// that are outside the 2xx range will nack the event, meaning it
// will be retried later.
w.WriteHeader(http.StatusOK)
_, _ = w.Write([]byte("OK"))
}
// verifyRequest verifies that the request originated from DT, and that
// the body hasn't been modified since it was sent. This is done by
// verifying that the checksum field of the JWT token matches the checksum
// of the request body, and that the JWT is signed with the signature secret.
func verifyRequest(bodyBytes []byte, tokenString string) error {
// Decode the JWT, and verify that it was signed using the
// signature secret. Also verifies the algorithm used to sign the JWT.
token, err := jwt.Parse(
tokenString,
func(token *jwt.Token) (interface{}, error) {
// Return out signature secret to verify that it was used to
// sign the JWT.
return []byte(signatureSecret), nil
},
jwt.WithValidMethods([]string{"HS256"}),
)
if err != nil {
return err
}
// Verify the request body checksum.
sha256Bytes := sha256.Sum256(bodyBytes)
sha256String := hex.EncodeToString(sha256Bytes[:])
claims := token.Claims.(jwt.MapClaims)
if sha256String != claims["checksum_sha256"] {
return fmt.Errorf("Checksum mismatch.")
}
return nil
}
// handleEvent processes the event itself. For this example,
// we will just decode a touch event, and print out the timestamp,
// device ID, and the device type.
func handleEvent(payload []byte) error {
// The structure of the events we'll received from a Data Connector.
type Event struct {
Event struct {
EventId string `json:"eventId"`
EventType string `json:"eventType"`
Data json.RawMessage `json:"data"`
Timestamp string `json:"timestamp"`
} `json:"event"`
Labels map[string]string `json:"labels"`
Metadata map[string]string `json:"metadata"`
}
// The structure of the `Event.Data` field for a touch event.
// We'll be using touch events for this example.
type TouchData struct {
Touch struct {
Timestamp string `json:"updateTime"`
} `json:"touch"`
}
// Decode the event
var event Event
if err := json.Unmarshal(payload, &event); err != nil {
return err
}
// First, check if the event type is one of the event
// types we're expecting.
// As an example, we'll check for touch events here.
switch event.Event.EventType {
case "touch":
// Now that we know this is a touch event, we can decode
// the `Event.Data` field.
var touchData TouchData
err := json.Unmarshal(event.Event.Data, &touchData)
if err != nil {
return err
}
// Also, since we now know this is a device event, we can
// check for the device type and device identifier
// in the event metadata.
deviceType := event.Metadata["deviceType"]
deviceId := event.Metadata["deviceId"]
timestamp := touchData.Touch.Timestamp
fmt.Printf("Received touch event at %s from %s sensor with id %s\n",
timestamp,
deviceType,
deviceId,
)
}
return nil
}
func main() {
// Sets up a bare-bones server that listens on port 8080, and
// routes all requests to the path "/" to the
// `DataConnectorEndpoint` function.
http.HandleFunc("/", DataConnectorEndpoint)
fmt.Println("Started listening on localhost:8080 ...")
if err := http.ListenAndServe(":8080", nil); err != nil {
fmt.Printf("Closed with error: %v\n", err)
} else {
fmt.Println("Closed server successfully")
}
}
Every event received by DT Cloud is put in a dedicated, per-Data Connector queue. Messages are removed from this queue once acknowledged, or if the message is older than 12 hours.
A side effect of this delivery guarantee is that, under certain conditions, you may receive duplicates of the same event. While rare, deduplication should be performed on the receiving end by checking event IDs.
Best Practice
Use the included eventId field to check for duplicated events.
Any time a Data Connector does not receive a successful response (HTTP status code 2xx), the event will be retried. If an event has not been successfully acknowledged after 12 hours, it will be discarded.
The retry interval is calculated as an exponential backoff policy, given by
where
is the initial interval of 8 seconds and
the retry counter. The interval will not exceed 1 hour. For very slow endpoints, the minimum retry interval will be
the response time.
Attempt | Retry Interval [s] |
1 | 8 |
2 | 16 |
3 | 32 |
... | ... |
9 | 2048 |
10 | 3600 |
11 | 3600 |