1. Uplink Configuration Structure

Kafka uplink is located at:

• config.uplink

Contains:

• enabled: Master switch (Default true)
• producer: Producer parameters (Default enables idempotence)
• Mapping by event type:
  • deviceConnected
  • deviceDisconnected
  • telemetry
  • attributes

Each mapping shape:

Field	Type	Description
enabled	boolean	Whether to enable this event type
topic	string	Topic template (Handlebars)
key	string	Key template (Handlebars), if rendered empty then key is not set
payload	object	Payload configuration (EnvelopeJson/Kv/TimeseriesRows/MappedJson)

Template syntax and variable table:

• Template Syntax (Handlebars)
• Template Variable Table

---

2. Default topic/key

• topic default: ng.uplink.{{event_kind}}.{{device_name}}
• key default: {{device_id}}

TIP

If you want to write to data lake partitioned by time, suggest putting time dimension into topic (yyyy/MM/dd/HH), and keep key partitioned by device.

---

3. Producer Parameters

Field	Default	Description
enableIdempotence	true	Whether to enable idempotent producer (Stronger delivery semantics)
acks	all	none/one/all
compression	lz4	none/gzip/snappy/lz4/zstd
lingerMs	5	Batch delay (ms)
batchNumMessages	1000	Max messages per batch
batchSizeBytes	131072	batch.size (128KiB)
messageTimeoutMs	30000	message.timeout.ms
requestTimeoutMs	10000	request.timeout.ms
maxInflight	5	max.in.flight.requests.per.connection

Ordering Hint

Even if idempotence is enabled, ordering semantics may still be affected by maxInflight when retries occur. If you are very sensitive to strict ordering, please understand Kafka's idempotence and in-flight semantics before adjusting these parameters.

---

4. Headers

Kafka uplink will write key-value pairs from RenderContext as headers (e.g., app_id, event_kind, device_id, etc.).

Usage:

• Platform side quick diagnosis and filtering
• Downlink filter.mode=property can utilize headers (Note: Only UTF-8 headers can be recognized)

---

5. Payload Selection

See: Uplink Payload Overview