Tools / Apache HertzBeat Interview Questions
Explain the internal working of HertzBeat's data collection pipeline?
The pipeline is a producer/consumer flow: the manager schedules a job, a collector executes it, and the resulting data fans out to multiple independent consumers over an internal message queue rather than being written directly to one destination.
flowchart LR
A[Manager schedules job] --> B[Collector executes protocol collect]
B --> C[Publish metrics to internal queue]
C --> D[Warehouse consumer: persist data]
C --> E[Alerter consumer: evaluate thresholds]
C --> F[Other consumers: e.g. OTEL/log pipeline]
Because collection publishes once to the queue and multiple consumers subscribe independently, adding a new downstream capability — like the alerter's threshold evaluation — doesn't require touching the collection code itself. The queue also decouples collection speed from storage speed, so a temporarily slow TSDB write doesn't stall the next collection cycle.
After a collector gathers data, it is published to:
Using a shared queue with independent consumers mainly allows:
More Related questions...
