Usage Metrics

Daytona measures every sandbox's resource consumption from outside the sandbox. Because sandboxes are hardware-isolated microVMs, the platform can observe CPU, memory, disk, and network usage from the host side, sampled roughly once per second — no agent runs inside your sandbox, so metrics collection adds zero overhead to your workload and can't be tampered with by anything running in it.

Three surfaces cover the common questions: a live endpoint for "what is this sandbox doing right now", a history endpoint for "what has it been doing over time", and a project report for "what is all of this costing me".

Live usage

GET /vms/:id/usage returns the sandbox's current resource usage — CPU, memory, disk, and network — as observed by the host. Use it for spot checks and for dashboards that watch a specific sandbox in real time.

curl https://api.rl.trydaytona.com/vms/$SANDBOX_ID/usage \
  -H "Authorization: Bearer $RLP_API_KEY"

The response wraps the live sample in a document with "source": "live". The sandbox must be running and placed; a sandbox with no compute (stopped, paused) has no live usage to report.

Usage history

GET /vms/:id/usage/history returns a time series of the same measurements, so you can see how a workload behaved over minutes or days — where the CPU spikes were, how memory grew, how much I/O a job did.

# Default: the last 24 hours at minute resolution
curl "https://api.rl.trydaytona.com/vms/$SANDBOX_ID/usage/history" \
  -H "Authorization: Bearer $RLP_API_KEY"
 
# A specific window at hourly resolution
curl "https://api.rl.trydaytona.com/vms/$SANDBOX_ID/usage/history?from=2026-07-30T00:00:00Z&to=2026-08-01T00:00:00Z&resolution=h" \
  -H "Authorization: Bearer $RLP_API_KEY"

Query parameters:

Each period in the response carries start, duration_ms, and integrals/gauges for the bucket: cpu_core_s (CPU core-seconds consumed), mem_byte_s, mem_rss_peak, disk_used, io_r_bytes, io_w_bytes, net_rx_bytes, and net_tx_bytes. To turn an integral into an average rate, divide by the bucket's duration — for example, average cores used = cpu_core_s / (duration_ms / 1000).

Project-level reporting

GET /usage/report aggregates usage across all sandboxes in your project over a time window, at hourly (resolution=h, default) or daily (resolution=d) granularity — the input for cost review and budgeting. See Platform → Spending for how usage maps to billing.

curl "https://api.rl.trydaytona.com/usage/report?start=2026-07-01T00:00:00Z&end=2026-08-01T00:00:00Z&resolution=d" \
  -H "Authorization: Bearer $RLP_API_KEY"

Dashboard

The Usage tab on a sandbox's page in the dashboard charts the same history data — the quickest way to eyeball a workload's footprint without writing any queries.

See also

• Platform → Spending — project-level cost breakdowns.
• Logs — when you need the why behind a usage spike.