Infrastructure monitoring
Investigate hosts, network devices, containers, and Kubernetes alongside the paths and services they support.
The application is slow, but its host, the switch, and the container platform each show a different part of the story. RouteSpear connects machine identity to the resources, interfaces, workloads, and real TCP connections that help you decide where to investigate next.
Follow the symptom into the systems behind it
- Hosts and processes
- Collect CPU, memory, filesystem, interface, and process measurements on Linux, Windows, and macOS. Use Remote Host Metrics over SSH for Linux systems where you do not install a persistent agent.
- Devices and interfaces
- Use SNMP and SNMP Discovery to observe network devices. Inspect utilization, errors, discards, neighbours, and learned addresses to connect an affected system to a device port.
- Containers and Kubernetes
- Follow workloads across reporting hosts, then inspect container resource use, limits, and throttling where supplied. Add Kubernetes API collection for replicas, pod scheduling, and node conditions.
- Kernel and connection evidence
- On supported Linux hosts, eBPF checks add scheduler wait, disk latency, and process events. Passive TCP observations add peer and process context, retransmits, and connection timing where supported.
Start with one slow service
Establish the affected place and time
Use the service test to identify the reporting agent and incident window. Find the service host in Infrastructure by name, address, or another reported identifier.
Inspect the next likely constraint
Compare host pressure with container limits, or inspect the device interface carrying the traffic. Add kernel or passive-flow collection when periodic resource snapshots leave the question open.
Keep the useful checks running
Add the supported metrics to a dashboard and configure health rules for the conditions that matter. Extend collection to the next system when it answers a specific operational question.
What the evidence can tell you
Collection determines coverage
Opening Infrastructure does not start a scan or enable a collector. Fields depend on the configured checks, platform, permissions, and successful samples. An unmeasured resource stays unknown.
Correlation needs context
A switch discard counter does not prove which application packets were dropped. A busy host does not establish a network fault. Keep each source, interface, reporting vantage, and sample time attached to the conclusion.
Bring the service that keeps sending you back to the host
Bring a recent incident or the environment you want to cover. We can walk through where to place agents and which measurements answer your questions.