Host Metrics

Sample every host in a swamp fleet — disk, network IO, load and memory — and push each sample to the swamp-managed OpenTelemetry system, every 15 minutes. WHAT IT DOES

The missing reporting half of a swamp fleet's host telemetry. It reuses @aaronge/host-health for the local disk (df), load, memory and thermal reads, and adds the two pieces host-health lacks: network IO counters (/proc/net/dev) and an OTLP push. listWorkers materialises the connected worker pool as ordinary data; a scheduled workflow fans a per-host sample out over every worker (dispatched to that worker, so it reads that host's /proc) plus the serve host itself, then emit resolves the OTLP endpoint and bearer token from the @svendowideit/otel-settings contract and POSTs one log record plus gauges — including the change in / free space since the previous sample. Prefer it over per-host scripts or a second metrics agent: the read methods already exist, the fleet list comes from swamp itself, and a host's whole report is one model instance.

INSTALL

swamp extension pull @svendowideit/host-metrics

DEPENDENCIES

@aaronge/host-health (the disk/load/memory/thermal reads this model consumes, pulled separately) and a rendered @svendowideit/otel-settings contract (the otel model) with a reachable OTLP gateway. Linux only (host-health is Linux-only). Without the contract, emit is a no-op that records why.

RUN

# Sample this machine once — network counter, then the fleet report.
swamp model @svendowideit/host-metrics method run readNetwork host-metrics-host
swamp model @svendowideit/host-metrics method run emit host-metrics-host \
  --input healthModel=host-health-host

# List the swamp worker pool as ordinary data (run on the orchestrator).
swamp model @svendowideit/host-metrics method run listWorkers host-metrics-fleet
swamp data query 'modelName == "host-metrics-fleet" && name == "workers"' --select content

# The scheduled pipeline: sample the serve host and every attached worker
# every 15 minutes and push each to OTel (run through `swamp serve`).
swamp workflow run @svendowideit/host-metrics-fleet --server ws://127.0.0.1:9090

CONFIGURE

Global arguments (set at model creation with --global-arg key=value):

  healthInstancePrefix  per-host host-health instance prefix  (host-health-)
  otelModel             @svendowideit/otel-settings model     (otel)
  otelVaultName         vault holding the OTLP bearer token   (empty)
  otelTokenRef          vault/env key for the bearer token    (OTEL_EXPORTER_OTLP_TOKEN)

Method arguments: readNetwork and listWorkers take none. emit takes host, healthModel, endpoint, mesh, otelModel, otelVaultName and otelTokenRef. The @svendowideit/host-metrics-fleet workflow takes otelModel, otelVaultName and otelTokenRef, and reads the worker list from listWorkers.

WHAT IT INSTALLS

Nothing on the host — no services, daemons, cron triggers or webhooks. It reads /proc and calls df through @aaronge/host-health, writes swamp resources, and POSTs to the collector. The schedule is a swamp trigger evaluated by swamp serve.