Skip to main content

Monitoring

Prometheus (metrics + rules), Alertmanager (notifications), and optional Alloy (collector) run on the host — not inside Grid Compose.

Stopping Grid (docker compose down, systemd stop) does not stop metrics.

ComponentPortRole
Prometheus:9090Store / query metrics; evaluate alert rules
Alertmanager:9093Route and notify firings
Alloy—Scrapes → remote_write to Prometheus

Install scripts and sample config live in grid-core/install/monitoring.

Production VM (Compose → host)​

sudo bash install/monitoring/install-systemd.sh
systemctl status prometheus alertmanager alloy

In install/.env:

GRID_PROMETHEUS_URL=http://host.docker.internal:9090
GRID_ALERTMANAGER_URL=http://host.docker.internal:9093
GRID_MONITORING_CONFIG_DIR=/etc/prometheus
GRID_MONITORING_HOST_DIR=/etc/prometheus

Then bring Grid up with bash install/compose-up.sh.

Local (npm run prod)​

GRID_PROMETHEUS_URL=http://127.0.0.1:9090
GRID_ALERTMANAGER_URL=http://127.0.0.1:9093
GRID_MONITORING_CONFIG_DIR=./install/monitoring

Start Prometheus with --web.enable-lifecycle so the console Settings page can POST /-/reload. Full commands: the install README linked above.

Console behavior​

  • Alerts — list firings from Alertmanager.
  • Settings — write rules/ + alertmanager.yml under GRID_MONITORING_CONFIG_DIR, then reload.
  • Collectors on managed infra — POST /api/v1/monitoring/setup/:id writes Alloy config that remote-writes to this host :9090.