A few things to consider at the receiving side when forwarding events with Data Connectors.
Request Contents
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.
Header
If a signature secret is set in the configuration, the following header will be included.
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.
Body
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.
The product number of the device that published the event.
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.
Acknowledging Received Event
A request-reply flow on the Endpoint URL should be implemented as follows:
Your endpoint receives an HTTPS POST request.
Your service processes the data in some way.
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.
Verifying Signed Events
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.
Extract the signed JWT from the HTTP header X-Dt-Signature of the received request.
Verify the JWT's signature with the signature secret.
Calculate a SHA256 checksum over the entire request body.
Compare the body checksum with the checksum_sha256 field contained in the JWT (there's a checksum field as well that uses SHA1 which is less secure than SHA256, and is kept only for backward compatibility).
If these checksums are identical, you can be certain that the event has not been tampered with and originated from your Data Connector.
# This Python script is built on Flask, docs are available here:# https://flask.palletsprojects.com/en/2.0.x/quickstart/import osimport hashlibimport jwt # pip install pyjwt==2.7.0from flask import Flask, request # pip install Flask==2.3.2app =Flask(__name__)# Read environment variable.SIGNATURE_SECRET = os.environ.get('DT_SIGNATURE_SECRET')@app.route('/', methods=["POST"])defdata_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.ifnotverify_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)defverify_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"])exceptExceptionas err:print(err)returnFalse# Verify the request body checksum. m = hashlib.sha256() m.update(body) checksum = m.digest().hex()if payload["checksum_sha256"]!= checksum:print('Checksum Mismatch')returnFalsereturnTruedefhandle_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("Got touch event at {} from {} sensor with id {}".format( timestamp, device_type, device_id, ))
# This Python is built on Flask, docs are available here:# https://flask.palletsprojects.com/en/2.0.x/quickstart/import os# pip install dtintegrations==0.5.1from dtintegrations import data_connector, providerfrom flask import Flask, request # pip install Flask==2.3.2app =Flask(__name__)# Read environment variable.DT_SIGNATURE_SECRET = os.getenv('DT_SIGNATURE_SECRET')@app.route('/', methods=['POST'])defdataconnector_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)defhandle_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 isNone: device_type = metadata.device_type device_id = metadata.device_id timestamp = payload.event.data.timestampprint("Got touch event at {} from {} sensor with id {}".format( timestamp, device_type, device_id, ))
constcrypto=require('crypto')constexpress=require('express') // npm install express@4.17.2constjwt=require('jsonwebtoken') // npm install jsonwebtoken@8.5.1// Read environment variableconstsignatureSecret=process.env.DT_SIGNATURE_SECRET// dataConnectorEndpoint receives, validates, and returns a response // for the forwarded event.constdataConnectorEndpoint= (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.constverifyRequest= (payload, token) => {// Decode the JWT, and verify that it was signed using the // signature secret. Also verifies that the algorithm used was HS256.let decodedtry { decoded =jwt.verify(token, signatureSecret, { algorithms:"HS256" }) } catch(err) {console.log(err)returnfalse }// Verify the request body checksum.consthash=crypto.createHash("sha256")constchecksum=hash.update(payload).digest("hex")if (checksum !==decoded.checksum_sha256) {console.log('Checksum Mismatch')returnfalse }returntrue}// 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.consthandleEvent= (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.constdeviceType=payload.metadata.deviceTypeconstdeviceId=payload.metadata.deviceIdconsttimestamp=payload.event.data.touch.updateTimeconsole.log(`Received touch event at ${timestamp} from ${deviceType} sensor with id ${deviceId}`)breakdefault:break }}// Sets up a bare-bones server that listens on port 8080, and routes// all requests to the path "/" to the `dataConnectorEndpoint` functionconstapp=express()app.use(express.json())app.listen(8080)app.post('/', dataConnectorEndpoint)
packagemainimport ("crypto/sha256""encoding/hex""encoding/json""fmt""io/ioutil""net/http""os"// go get github.com/golang-jwt/jwt/v5@v5.0.0 jwt "github.com/golang-jwt/jwt/v5")// Read environment variablevar signatureSecret = os.Getenv("DT_SIGNATURE_SECRET")// DataConnectorEndpoint receives, validates, and returns a response// for the forwarded event.funcDataConnectorEndpoint(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.funcverifyRequest(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.") }returnnil}// 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.funchandleEvent(payload []byte) error {// The structure of the events we'll received from a Data Connector.typeEventstruct { 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.typeTouchDatastruct { Touch struct { Timestamp string`json:"updateTime"` } `json:"touch"` }// Decode the eventvar event Eventif 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, ) }returnnil}funcmain() {// 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") }}
Handling Duplicates
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.
Retry policy
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
t0⋅2n−1,
where t0 is the initial interval of 8 seconds and n the retry counter. The interval will not exceed 1 hour. For very slow endpoints, the minimum retry interval will be 4x the response time.