Apache Pulsar¶

The pulsar probe monitors an Apache Pulsar broker via its Admin REST API (/admin/v2/brokers/ready) and Prometheus metrics endpoint (/metrics), reporting broker health, throughput, storage and backlog at the broker level.

Quick start¶

# probes.d/10-pulsar.yaml — each file under probes.d/ is a YAML array of probes
- name: pulsar
  type: pulsar
  params:
    endpoint: http://localhost:8080

Parameters¶

Parameter	Must set	Default	Description
`endpoint`	In practice	`http://localhost:8080`	Base URL of the broker admin and metrics HTTP service. Example: `http://pulsar.example.com:8080`
`timeout`	No	`10`	Request timeout in seconds
`interval`	No	`60`	Seconds between collections
`instance_name`	No	-	Stable identity of this broker instead of the one derived from the endpoint

Metrics¶

Metric	Unit	Description
`senhub.pulsar.up`	1	1 when the broker answered `/admin/v2/brokers/ready` with HTTP 200
`pulsar.topics.count`	{topic}	Number of topics on the broker
`pulsar.rate.messages.in`	{message}/s	Incoming message rate (broker-level aggregate)
`pulsar.rate.messages.out`	{message}/s	Outgoing message rate
`pulsar.throughput.in`	By/s	Incoming byte throughput
`pulsar.throughput.out`	By/s	Outgoing byte throughput
`pulsar.storage.size`	By	Broker-level storage used for ledger data
`pulsar.message.backlog`	{message}	Total message backlog across all topics
`pulsar.producers.count`	{producer}	Connected producers
`pulsar.consumers.count`	{consumer}	Connected consumers

Operational notes¶

The Pulsar admin port (default 8080) and metrics endpoint are on the same port.
For TLS-secured brokers, use https:// as the endpoint prefix.
The probe reads the broker-level aggregates from /metrics (the pulsar_broker_* series), which a broker publishes even before any topic exists. A broker too old to publish them is read through its per-namespace series instead, one per namespace. Per-topic series are not read.

Metric reference¶

Every metric this probe can emit. Metric is the OpenTelemetry name the OTLP, Prometheus and Zabbix outputs derive theirs from. Name is what a Nagios check and the API metrics= filter match. PRTG channel is the label PRTG shows, placeholders filled from the series' tags.

Metric	Name	PRTG channel	Unit	Description
`senhub.pulsar.up`	`senhub.pulsar.up`	Pulsar {endpoint} Up	#	1 when the broker answered /admin/v2/brokers/ready with HTTP 200
`pulsar.topics.count`	`pulsar.topics.count`	Pulsar Topics	#	Number of topics on the broker
`pulsar.producers.count`	`pulsar.producers.count`	Pulsar Producers	#	Number of producers connected to the broker
`pulsar.consumers.count`	`pulsar.consumers.count`	Pulsar Consumers	#	Number of consumers connected to the broker
`pulsar.rate.messages.in`	`pulsar.rate.messages.in`	Pulsar Publish Rate	msgs/s	Messages published per second to the broker
`pulsar.rate.messages.out`	`pulsar.rate.messages.out`	Pulsar Dispatch Rate	msgs/s	Messages dispatched per second from the broker
`pulsar.throughput.in`	`pulsar.throughput.in`	Pulsar Publish Throughput	Bytes/s	Bytes published per second to the broker
`pulsar.throughput.out`	`pulsar.throughput.out`	Pulsar Dispatch Throughput	Bytes/s	Bytes dispatched per second from the broker
`pulsar.storage.size`	`pulsar.storage.size`	Pulsar Storage Size	B	Total storage used by the broker for message data
`pulsar.message.backlog`	`pulsar.message.backlog`	Pulsar Message Backlog	#	Number of messages in the backlog (not yet consumed)
`pulsar.storage.read.rate`	`pulsar.storage.read.rate`	Pulsar Storage Read Rate	entries/s	Storage ledger entries read per second
`pulsar.storage.write.rate`	`pulsar.storage.write.rate`	Pulsar Storage Write Rate	entries/s	Storage ledger entries written per second