Skip to content
Eclipse Sparkplug 3.0.0 · ISO/IEC 20237:2023

Sparkplug B

Sparkplug B standardizes industrial messaging over MQTT. It defines a topic namespace, typed Protobuf payloads and operational state management, so edge nodes, devices and host applications exchange data with consistent lifecycle awareness.

Category
Application and messaging profile
Built on
MQTT 3.1.1 over TCP, TLS or WebSocket
Topic namespace
spBv1.0
Library
go-sparkplugb
CLI
sparkplugbctl
Interop
sparkplugb-interop v0.1.0

Sparkplug B is not a replacement for MQTT, and it is not an equipment model. It is a profile of MQTT: it says how industrial devices and applications use an MQTT broker so that they understand each other without per-project agreements. Plain MQTT moves bytes on topics that anyone can invent. Sparkplug fixes three things on top of it:

  1. A topic namespace: spBv1.0/<group>/<message type>/<edge node>/[<device>], plus spBv1.0/STATE/<host id> for host applications.
  2. A payload: a Protobuf message with named, typed metrics, timestamps, aliases, DataSets and Templates.
  3. Session state: birth and death certificates for every edge node and device, sequence numbers, and host awareness, so a consumer always knows whether a value is current or stale.

Unlike SunSpec, Sparkplug B does not standardize the meaning of equipment points such as inverter power or a battery’s state of charge. It standardizes how named, typed metrics and the state of the devices behind them are represented and exchanged reliably across an MQTT infrastructure. A Sparkplug metric can carry a SunSpec or IEC 61850 value; Sparkplug itself does not say what the value means.

Roles and messages

Role What it is Publishes Receives
Edge node A gateway, PLC or controller that speaks MQTT for itself and the devices behind it NBIRTH, NDATA, NDEATH (as its MQTT will), DBIRTH, DDATA, DDEATH NCMD, DCMD, STATE of its primary host
Device Equipment behind an edge node, with no MQTT connection of its own through its edge node through its edge node
Primary host application The SCADA or MES system that edge nodes wait for before publishing; it tracks every node and device STATE (online and offline), NCMD, DCMD everything under spBv1.0/
Application Any other consumer: a dashboard, historian or analytics pipeline NCMD, DCMD if it needs to write everything under spBv1.0/

How the session state works

  • Birth certificates. An edge node starts every session with an NBIRTH that lists every metric with its name, datatype, current value and optional alias, then one DBIRTH per device. Later NDATA and DDATA messages report by exception: only what changed, often by alias instead of by name.
  • Death certificates. The edge node registers its NDEATH as the MQTT will when it connects, so the broker publishes it if the connection drops. A bdSeq number in both the NBIRTH and the will ties a death to the session it ends, so a late will of an old session is not mistaken for the current one.
  • Sequence numbers. Every message of a node session carries a seq from 0 to 255 that wraps. A host that sees a gap can wait briefly for the missing message and then ask the node to start over with a rebirth, an NCMD that writes Node Control/Rebirth.
  • Primary host awareness. A primary host application publishes a retained STATE message, online or offline, on spBv1.0/STATE/<host id>. An edge node configured for it waits for that host to be online before its births, and when the host goes offline it ends its session and walks to the next MQTT server, so data is never published to a broker the host cannot see.
  • Commands. NCMD and DCMD write metrics on a node or device, from setpoints to the rebirth request.

Where you meet it

  • Unified namespace and IIoT architectures that bring PLC, RTU and gateway data to SCADA, MES, historians and cloud platforms through one MQTT broker
  • Ignition and other SCADA platforms with MQTT modules, and edge gateways and controllers that publish their tags as Sparkplug edge nodes
  • Brokers that are Sparkplug aware, keeping the latest birth certificates for consumers that connect late
  • Energy, water, oil and gas and manufacturing sites where many remote assets report by exception over cellular or satellite links

Quickstart

From the command line

Download sparkplugbctl from GitHub Releases, or go install github.com/otfabric/sparkplugbctl@latest. With an MQTT broker on localhost:1883, start a primary host application, an edge node that waits for it, and a listener, each in its own terminal:

Terminal window
sparkplugbctl host --host-id scada
sparkplugbctl edge --group plant --node gw1 --device sensor1 \
  --metric "temperature:Double=22.5" --interval 2s --primary-host scada
sparkplugbctl listen --group plant

listen decodes every message on spBv1.0/#, STATE included:

[NBIRTH] spBv1.0/plant/NBIRTH/gw1
  seq=0  ts=2026-10-27T11:12:18Z
  bdSeq                 Int64    1
  Node Control/Rebirth  Boolean  false
  temperature           Double   22.5

From Go

Terminal window
go get github.com/otfabric/go-sparkplugb

An edge node that publishes its birth certificate and then reports by exception (from the go-sparkplugb README):

main.go
transport := mqtt.NewPahoTransport()


node, _ := edge.New(transport, edge.Config{
  GroupID:    "plant",
  EdgeNodeID: "gateway-01",
  BrokerURL:  "tcp://broker:1883",
})


ctx := context.Background()
node.Connect(ctx)
defer node.Close(ctx)


temp, _ := payload.NewMetric("temperature", pb.DataType_Double, float64(22.5))
node.PublishBirth(ctx, []*pb.Payload_Metric{temp})


updated, _ := payload.NewMetric("temperature", pb.DataType_Double, float64(23.1))
node.PublishNodeData(ctx, []*pb.Payload_Metric{updated})

What go-sparkplugb supports

go-sparkplugb is pre-release (v0.1.x). Its own conformance matrix maps the specification clause by clause; this is a summary.
Roles Edge node with devices, primary host application, and application without STATE
Session state NBIRTH, DBIRTH, NDEATH as the will with bdSeq, DDEATH, rebirth on request, births that wait for the primary host’s STATE
High availability Edge node server walking across several MQTT servers when the primary host goes offline; the host takes several server URLs but connects to one at a time
Host behaviour seq checks, messages held and applied in order when a broker reorders them, one pending rebirth request per node, NDEATH of an earlier session ignored
Payloads Every scalar and array datatype, DataSets, Templates (definitions and instances), property sets, quality codes, metric aliases
Topic namespace Parser and builder for spBv1.0/... with identifier validation
Transport MQTT 3.1.1 over TCP, TLS or WebSocket through Eclipse Paho; other MQTT clients through the mqtt.Transport interface
Not covered MQTT 5 (and its DISCONNECT with will), checks that NDATA and DDATA only use metrics of the birth, the SendComplexData group of the edge conformance profile
Go version 1.26 or later

Interoperability, not certification

sparkplugb-interop packages six independent Sparkplug B implementations as edge nodes and host applications in containers: Eclipse Tahu in Java and JavaScript, pysparkplug (Python), SparkplugNet (C#), srad (Rust) and sparkplughost (Go). On every push, go-sparkplugb’s host and application run against each reference edge node, and its edge node against each reference host, on Eclipse Mosquitto, HiveMQ CE, EMQX and NanoMQ. The steps cover births, aliases, updates of all fourteen scalar datatypes, NCMD and DCMD writes, rebirth, device death and birth, NDEATH, the will after a crash, and the primary host going offline. Where a reference lacks a capability or has a known upstream bug, the step is skipped and the reason logged.

This is interoperability testing, not the Sparkplug Compatible program. That program requires passing the official Eclipse Sparkplug TCK for a profile and being listed by the Eclipse Foundation. go-sparkplugb has not been run against the TCK and does not claim Sparkplug Compatible status. Templates, DataSets, arrays and properties are covered by the library’s own codec tests only, because no reference fixture uses them yet.

Standards and licensing

The Sparkplug specification is developed by the Eclipse Sparkplug Working Group and was adopted as ISO/IEC 20237:2023 through ISO/IEC JTC 1. Anyone can implement it without fee or royalty; the specification text is under the Eclipse Foundation Specification License. Eclipse Tahu is the Working Group’s open-source implementation, and the Eclipse Sparkplug TCK tests implementations for the Sparkplug Compatible listing. “Sparkplug” is a trademark of the Eclipse Foundation.

go-sparkplugb is MIT-licensed, except its proto/ directory: the canonical sparkplug_b.proto from Eclipse Tahu and the Go code generated from it stay under EPL-2.0, a weak copyleft licence. If you change those files and distribute them, the EPL-2.0 terms apply to your changes. sparkplugbctl and sparkplugb-interop are MIT-licensed; the interop images also contain the upstream stacks under their own licences (EPL-2.0, Apache-2.0, MIT and others).

OT Fabric projects for Sparkplug B

go-sparkplugb

Sparkplug B 3.0.0 over MQTT for Go: edge node, primary host application and application roles, with birth and death certificates, primary host awareness, server walking, aliases, sequence checks and the full typed metric codec.

go get github.com/otfabric/go-sparkplugb

sparkplugbctl

Watch and decode Sparkplug B traffic on a broker, publish typed metrics, encode and decode payloads, or simulate an edge node, a primary host application or an application.

sparkplugb-interop

Six Sparkplug B reference stacks in six languages, Eclipse Tahu (Java and JavaScript), pysparkplug, SparkplugNet, srad and sparkplughost, as edge nodes and host applications on Mosquitto, HiveMQ CE, EMQX and NanoMQ.

Who is behind Sparkplug B

The organizations that publish, test, certify or host open-source work for Sparkplug B. Listed for reference, not as partners. Compare membership and fees.