⛏️GetSplunky!
Lv.1Intern
0 XP
0
Home Map Stats Drill Board Log In
CoursesArticlesSPL Cheat SheetCertification PathAboutPrivacyTerms

Splunky is an independent, free study resource. It is not affiliated with, endorsed by, or sponsored by Splunk or Cisco. Splunk, SPL and related marks are trademarks of Splunk, a Cisco company, and are used here only to identify the certifications this site helps you study for.

© 2026 Splunky

Back to Floor
The Telemetry Pipeline • Room 2

Agent vs. Gateway Deployment Modes

The Splunk OTel Collector runs in two complementary topologies: local Agent mode (DaemonSet or host) and centralized Gateway mode.

In Agent mode (agent_config.yaml), one Collector runs on every Linux/Windows host or as a Kubernetes DaemonSet pod per node. Because it lives locally, the Agent scrapes hostmetrics, kubeletstats, and local application OTLP traffic over loopback, attaching node-specific metadata via resourcedetection.

In Gateway mode (gateway_config.yaml), Collectors run as a scalable Kubernetes Deployment behind a load balancer. Gateways do not scrape local host metrics; instead, they receive aggregated OTLP streams from multiple Agents, apply stateful or heavy transformations (such as attribute scrubbing, filtering, or trace tail_sampling), and centralize Splunk access tokens and outbound firewall egress.

Agent -> Gateway -> Splunk O11y Cloud Topology
[Host Apps + Node] -> OTel Agent (DaemonSet / systemd)
                         | (otlp gRPC :4317)
                         v
                    OTel Gateway (Deployment behind LB)
                         | (signalfx exporter + access token)
                         v
                    Splunk Observability Cloud Ingest
On the SPLK-4001 exam, remember: hostmetrics and kubeletstats belong on the Agent (DaemonSet), while centralized token management, heavy filtering, and tail sampling belong on the Gateway.

Knowledge Check

Prove your understanding to clear the room (Rewards XP)
❤️❤️❤️
Question 1 of 2
A platform team wants to scrape CPU and disk metrics from 200 Kubernetes worker nodes while keeping Splunk ingest tokens and outbound HTTPS egress centralized in a single tier. Which architecture fits best?
AGateway-only Deployment configured to scrape hostmetrics from remote nodes over SSH
BAgent-only DaemonSet on every node configured without any outbound exporters
CAgent DaemonSet per node forwarding via OTLP to a central Gateway Deployment
DInstall the Prometheus server on every container alongside the application process