sysinfo — System Info & Benchmark¶

The sysinfo plugin (bundle io.cresco.sysinfo) gathers live operating-environment metrics for an agent — CPU, memory, disk, filesystem, and network — via OSHI, and can run an on-demand SciMark2 CPU benchmark. It is the resource-information provider the controller's PerfControllerMonitor queries.

At a glance¶

Fact	Value
Module path	`code/sysinfo`
Bundle symbolic name	`io.cresco.sysinfo`
Java files	6
Loaded	At runtime by the controller's `StaticPluginLoader`.
Type	Functional plugin (`@Component` implementing `PluginService`).
Metric source	OSHI (system) + SciMark2 (CPU benchmark).

Key classes¶

Class	Role
`ExecutorImpl`	Handles the EXEC actions (below).
`SysInfoBuilder`	`getSysInfoMap()` assembles `os` / `cpu` / `mem` / `disk` / `fs` / `part` / `net` JSON from OSHI. `getCPUInfo` sleeps ~1s for a load delta; the `fs` section is skipped on Raspbian.
`PerfSysMonitor`	Continuous sampler — a `Timer` that publishes a live-metrics `MapMessage` to `TopicType.AGENT` on the data plane every `perftimer` ms.
`Benchmark`	`bench()` runs the five SciMark2 kernels (FFT, SOR, Monte Carlo, Sparse Matmult, LU); `getJSON()` flattens the result.
`BenchMetric`	Immutable benchmark-result holder.

Executor actions¶

ExecutorImpl handles these EXEC actions (MsgEvent type EXEC):

Action	Behaviour
`getsysinfo`	Return a compressed `perf` param — the live `SysInfoBuilder.getSysInfoMap()`. This is what `PerfControllerMonitor` RPCs for resource info.
`getbenchmark`	Return a compressed `bench` param — `Benchmark.getJSON()`, running SciMark2 on demand.

Data-plane publishing and the on-demand benchmark¶

When enable_perf is set, PerfSysMonitor streams a live-metrics message to TopicType.AGENT every perftimer ms, feeding the data plane and, through it, dashboards and CEP.

SciMark2 is on-demand only

The SciMark2 benchmark is deliberately not run in the continuous sampler. It was previously burned at startup and bundled into every heartbeat; it is now executed only when a getbenchmark action is received. This keeps steady-state telemetry cheap. See Metrics & Measurements.

Configuration read¶

Key	Default	Meaning
`enable_perf`	`false`	Run the continuous live-metrics sampler.
`perftimer`	`10000`	Sampler interval (ms).
`pluginname`	—	This plugin's name.

See Configuration.

Where it fits¶

The controller's PerfControllerMonitor serves on-demand region/agent/plugin resource queries by RPCing getsysinfo to io.cresco.sysinfo. The plugin is therefore the authoritative source of a host's resource picture, both pushed (via the data-plane sampler) and pulled (via the controller's getMetrics/getSysInfo path).

Observability¶

Metrics — getmetrics exposes 30 MeasurementEngine gauges across the processor, memory, disk, network, sensors, power, system, gpu, and devices groups (CPU load, RAM, disk/NIC I/O, temperature, battery, uptime, and more), aggregated by getmetricinventory. Every gauge, its OSHI source, and its meaning is catalogued in Metrics & Measurements › sysinfo.
Health — registers the sysinfo Felix HealthCheck (tag local): OK while host telemetry is being served. See Health & State.