Arc, monitoring itself.
CPU, memory, network and disk for the container running every demo on this site, collected every 10 seconds and stored in Arc.
Spikes line up with demo traffic: vessel, flight and weather ingestion, plus the queries from people viewing the demos right now.
How it works
Collection: a Python script in its own container reads the Docker Engine API through a read-only /var/run/docker.sock every 10 seconds.
Ingestion: each sample is written to Arc with the MessagePack columnar format.
Metrics: CPU percent, memory usage and limit, and cumulative network and disk byte counters. The chart turns the counters into per-second rates.
Storage: everything lands in Arc's system_monitoring.container_metrics table as Parquet.
Summarize it in SQL
SELECT container_name, AVG(cpu_percent) AS avg_cpu, AVG(memory_percent) AS avg_memory, MAX(memory_usage_mb) AS peak_memory_mb FROM system_monitoring.container_metrics WHERE time > NOW() - INTERVAL '1 hour' AND container_name = 'arc-demo' GROUP BY container_name;