Tower and continuous metrics

A Tower is a small gateway that runs on a machine in your network. Server and database Collectors (or your own OpenTelemetry Collector) send their metrics to the Tower over OTLP/HTTP; the Tower forwards them to the controller over one outgoing link. The controller keeps metrics in its own PostgreSQL, next to your run results.

• The Tower only connects to the controller. Nothing connects into your network and no internet access is needed.
• If it cannot reach the controller it keeps incoming metrics on disk (1 GB / 24 hours by default) and forwards them in order once the link is back. When the buffer is full the oldest go first and are counted.
• One Tower per network zone (data centre, DMZ, cloud VPC) is enough.

Page: Towers & metrics In Spitfire: /towers (install admins only).

#Turning metrics on

1. On Towers & metrics, switch on Collect and keep metrics.
2. Tick the box confirming that metrics take space in the database. Who turned it on and when is recorded.
3. Enter the Retention. It cannot exceed your license's ceiling:

4. Enter the Storage budget (GB). You are warned at 80%; at 90% no new metrics are written.
5. Save.

#Installing a Tower

1. Add Tower → a name (e.g. istanbul-dc) and an optional zone.
2. Two tokens are shown only once; keep both:
   • sftower_…: the Tower's own token (SPITFIRE_TOWER_TOKEN).
   • sfcol_…: the token of the Collectors sending to the Tower (Authorization: Bearer sfcol_…).
3. Run the Docker command shown on the Tower machine:

docker run -d --name spitfire-tower --restart unless-stopped -p 4318:4318 \
  -v spitfire-tower:/var/lib/spitfire-tower \
  -e SPITFIRE_TOWER_CONTROLLER=https://<controller> \
  -e SPITFIRE_TOWER_TOKEN=sftower_... \
  <spitfire image> spitfire-tower

4. The list shows Connected (the first link takes a few seconds).

Tower settings (environment variables):

#Server and PostgreSQL metrics (Collector)

Spitfire writes a ready config for the official OpenTelemetry Collector Contrib (otel/opentelemetry-collector-contrib:0.164.0). Install the Collector on the server to watch; as long as it can reach PostgreSQL over the network, the same Collector reads the database too.

1. Towers & metrics → Collector config.
2. Tower address: http://localhost:4318 when the Tower is on the same machine, otherwise http://<tower-machine>:4318.
3. This server's metrics: CPU, memory, disk, network, load, file systems.
4. Add PostgreSQL: address (db.corp:5432), user (default spitfire_monitor), databases (empty: all), TLS.
5. Pick Docker or the otelcol-contrib binary → Generate config.
6. Save collector.yaml on the Collector machine.
7. Create the monitoring user in PostgreSQL as an admin:

CREATE ROLE spitfire_monitor WITH LOGIN PASSWORD '<password>';
GRANT pg_monitor TO spitfire_monitor;

8. Fill in the env and run the given command:

docker run -d --name spitfire-collector --restart unless-stopped --network host --pid host \
  -v /:/hostfs:ro -v "$PWD/collector.yaml:/etc/otelcol-contrib/config.yaml:ro" \
  -e SPITFIRE_COLLECTOR_TOKEN=<sfcol_…> \
  -e SPITFIRE_PG_PASSWORD=<password> \
  otel/opentelemetry-collector-contrib:0.164.0

9. Within a minute the server and the database show in the Sources list.

#In the run report

Every server and database that sends metrics while a test run is going shows on the run's Backend tab as the Tower metrics source; the test needs no connection for it:

• Server: CPU %, memory %, load (1 min), disk I/O, network I/O.
• Database: connections, connections / max %, commits/s, rollbacks/s, deadlocks/s, row operations/s.

Each value is a stage (ramp, plateau…) mean in the table and a chart over time.

#Connecting your own Collector

Add an exporter to your OpenTelemetry Collector's settings (replace <tower> with the Tower machine's address):

exporters:
  otlphttp/spitfire:
    endpoint: http://<tower>:4318
    headers:
      Authorization: "Bearer sfcol_..."
service:
  pipelines:
    metrics:
      exporters: [otlphttp/spitfire]

Once the Collector restarts, the source shows in the Sources list. Servers are counted by host.name, databases by server.address[:port] (else host.name).

#License limits

Your plan sets how many servers and databases you can take metrics from:

Growth and Scale can add packs: +10 servers or +5 databases (License page → add-on packs). At the limit new sources are not taken and are listed on the page. Delete a source you no longer use from Sources to make room; if its Collector keeps sending, it comes back.

#Troubleshooting