Splunk Processing Components

Splunk groups its primary runtime roles into processing components and management components. Processing components are the parts that ingest, store, search, and coordinate data flow through a deployment.

Component Types

ComponentCore jobTypical placementVariants
ForwarderIngest raw data and forward it onwardUsually on application hosts such as web serversUniversal forwarder, heavy forwarder, light forwarder, intermediate forwarder
IndexerReceive, index, store, and search dataUsually on dedicated machinesIndependent indexer, clustered indexer / peer node
Search headHandle user search requests, distribute searches to indexers, merge results, expose dashboards/toolsUsually on dedicated machinesIndependent search head, search head cluster member, search head node in an indexer cluster, search head pool member

Forwarders

Forwarders are the entry point for external data. The universal forwarder is the most common type. In more complex deployments, forwarders can send data through intermediate forwarders before it reaches indexers.

Key implications:

  • Data input is distributed close to the source systems
  • Forwarders are expected on non-Splunk application hosts
  • Not every forwarder is a full Splunk Enterprise instance; the universal forwarder is a lightweight separate executable

Indexers

Indexers store the data and participate in search execution. In non-clustered deployments they are independent nodes. In clustered deployments they are peer nodes that replicate data among themselves for availability.

The topology page treats indexers as the middle processing tier:

  • receive incoming data from forwarders
  • index and store that data
  • execute search work on local data when search heads dispatch requests

Search Heads

Search heads manage the search experience rather than storing the data themselves. They receive user requests, distribute search work to indexers, merge results, and provide operational and analytical interfaces such as dashboards.

The page notes one secondary behavior worth remembering: a search head can forward its own internal data to indexers, but that forwarding is not its primary role.

Operational Rules of Thumb

  • A deployment usually contains only some of the possible component types
  • An instance usually hosts at most one primary processing component
  • A single machine can host multiple instances, but Splunk describes that as unusual outside test environments

See Also