How to Build an Effective Nginx Monitoring System for E‑Commerce

Background and Nginx Basics

Nginx is an open‑source, high‑performance HTTP and reverse‑proxy server that can also serve IMAP/POP3. Its asynchronous, non‑blocking architecture gives it high concurrency, low resource consumption, and strong extensibility, making it well‑suited for e‑commerce workloads.

Key Monitoring Metrics

The full request lifecycle is logged in access.log and error.log. Critical metrics to monitor include:

Request latency ( $request_time, $upstream_response_time)

HTTP error codes (5xx, 4xx, redirects)

Traffic volume (PV, QPS)

Server saturation (CPU, worker connections, network I/O)

Monitoring Practices

Latency

Track $request_time and plot the TP99 value; optionally monitor $upstream_response_time to pinpoint upstream delays. Typical results show 90% of requests under 0.1 s and 99% under 0.3 s.

Error Monitoring

Monitor both service availability (semantic health check via HTTP request) and error codes. Add alerts for frequent 5xx responses and watch 4xx for permission or resource issues. Visualize status‑code distribution.

Traffic Monitoring

Observe total request count, detect spikes or drops, and set alerts for abnormal variations (e.g., >20% change). Track per‑minute request volume and correlate with business cycles.

Saturation

Measure resource utilization (CPU, network I/O, disk) and worker process limits ( worker_processes × worker_connections). Use http_stub_status_module to expose real‑time stats.

Visualization Solutions

Two open‑source stacks are presented:

ELK (Elasticsearch + Logstash + Kibana) : Real‑time log indexing, custom Nginx filters, and dashboards for latency, errors, traffic, and saturation.

Grafana + Elasticsearch + Rsyslog : Grafana provides flexible charting; combined with Elasticsearch it offers powerful log search and visualization.

Real‑World Cases

Case 1 – Traffic Surge

During a product flash‑sale, PV spiked dramatically, causing increased request latency. Alerts on PV and request time triggered investigation; the ELK dashboard revealed many bot IPs. Rate‑limiting and higher‑level DDoS protection were applied, stabilizing the service.

Case 2 – Error‑Code Alert

An upstream misconfiguration caused a surge of 500 and 302 responses. ELK filtering of 302 URLs identified the faulty module, and the upstream target was corrected. Subsequent upgrades to OpenResty with Lua health checks prevented recurrence.

Case 3 – Disk Exhaustion

A sudden increase in image upload size filled the client_body_temp_path partition, leading to 500 errors. The error log showed “No space left on device”. The root cause was a promotion that boosted upload volume; the fix involved expanding disk space and adjusting upload limits.

Conclusion

Effective Nginx monitoring combines latency, error, traffic, and saturation metrics with visual dashboards. Using ELK or Grafana provides flexibility for log search and real‑time charts, enabling rapid detection and resolution of performance issues in high‑traffic e‑commerce environments.