Skip to content
Tutorials Center
My learning

Topics

Lesson 6 of 36 ยท Architecture and Setup

Splunk Architecture: Forwarders, Indexers and Search Heads

Understand the Splunk data pipeline and the main components: universal and heavy forwarders, indexers, search heads, deployment server, cluster manager and license manager.

A single-machine Splunk does everything: it collects, indexes and searches. At scale those jobs are split across machines with different roles. Learning the roles first makes every later topic (forwarders, clustering, scaling, troubleshooting) easier to place.

The three tiers

Splunk's work divides into three stages:

  1. Data input tier: collects data and sends it on. Components: forwarders.
  2. Indexing tier: parses, indexes and stores events, and answers search requests for its own data. Components: indexers.
  3. Search management tier: where users log in, write searches and share dashboards. Components: search heads.
flowchart LR
    subgraph Sources
      A["Web servers"]
      B["App servers"]
      C["Network devices"]
    end
    A --> UF1["Universal<br/>Forwarder"]
    B --> UF2["Universal<br/>Forwarder"]
    C --> HF["Heavy<br/>Forwarder"]
    UF1 --> IDX1["Indexer 1"]
    UF2 --> IDX2["Indexer 2"]
    HF --> IDX1
    SH["Search head"] -->|"distributes searches"| IDX1
    SH -->|"distributes searches"| IDX2
    U["Users (browser)"] --> SH

Forwarders send data to indexers; search heads ask all indexers and merge the answers.

Forwarders

A forwarder is a Splunk instance that sends data to another instance. Two kinds:

Universal Forwarder (UF) Heavy Forwarder (HF)
Size Tiny agent, minimal CPU/RAM Full Splunk Enterprise install
Parses data No (sends raw streams, with a few exceptions) Yes, can parse, filter and route
Web UI None Yes (optional)
Typical use Install on every host Collect from APIs, filter or mask data before it travels

Use the UF by default. Use an HF when you need to parse, mask, or route events before they reach the indexer, or when a modular input (cloud or database collector) needs a full Splunk to run.

Forwarders also provide load balancing (spread data across several indexers), automatic failover, optional compression and SSL, and an indexer acknowledgment option so data is not lost on the way.

Indexers

An indexer receives data (usually on port 9997), runs the parsing pipeline (line breaking, timestamps, extractions at index time), writes events to disk and answers searches against the data it stores. Index storage uses buckets that age from hot to warm to cold to frozen, covered in indexes, buckets and retention.

Several indexers can be joined into an indexer cluster that keeps copies of every bucket so a machine can fail without data loss.

Search heads

A search head hosts Splunk Web, holds the knowledge objects (saved searches, dashboards, field extractions, lookups) and sends the search to the indexers (called search peers). Indexers return partial results; the search head merges them into the final answer. A search head cluster shares configuration among several search heads for availability and more users.

Management components

Component Job
Deployment server Pushes apps and configuration to many forwarders; forwarders are deployment clients
Cluster manager (older name: cluster master) Coordinates an indexer cluster: replication, which copies live where
Search head cluster deployer Distributes apps to search head cluster members
License manager Holds the license pool; other instances report usage to it
Monitoring Console Dashboards showing the health of the deployment
Python Stand-in: which component does the job?

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The data pipeline inside Splunk

Every event passes four stages:

  1. Input: a forwarder or the indexer reads the raw stream.
  2. Parsing: break the stream into events, set the timestamp and metadata, apply index-time rules (filtering, masking, routing).
  3. Indexing: write events and the search index to disk.
  4. Search: apply search-time knowledge (field extractions, lookups, calculated fields) when a user runs a search.

This split explains a key rule: parsing happens where data is first handled by a full Splunk instance (an HF if you have one, otherwise the indexer), not on a universal forwarder. Configuration for parsing must live on that tier; search-time settings belong on the search head.

Python Stand-in: events fanning out across indexers (auto load balancing)

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The same idea is why spreading data evenly matters: a search is as fast as its slowest indexer.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which component hosts Splunk Web and sends searches to the indexers?

    The search head is the user-facing tier that distributes searches and merges results.
  2. 2. Which forwarder can parse and filter events before sending them on?

    A heavy forwarder is a full Splunk Enterprise instance and can parse, filter and route.
  3. 3. What does a deployment server do?

    Deployment clients poll the deployment server for apps and configuration updates.
  4. 4. In which pipeline stage are field extractions normally applied?

    Most extractions are search-time; only a few settings (timestamp, line breaking, routing) are index-time.

Summary

  • Forwarders collect, indexers store and parse, search heads search.
  • Use universal forwarders everywhere; heavy forwarders only when parsing or routing before the indexer is needed.
  • Deployment server, cluster manager, license manager and Monitoring Console manage the platform.
  • Parsing configuration belongs on the first full Splunk instance, search-time configuration on the search head.

Next: Deployment topologies and sizing