Claude
d40c30d0e8
- Web Server reads static files inline; proxy middleware chain flattened - About window with license summary and project links; AGPL-3.0-or-later - Rewritten documentation: feature list, screenshots, performance guide - Fix false spikes in per-Vhost request-rate charts Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MMsgSXoK4Bph2vU6fSBkAH
52 lines
2.9 KiB
Markdown
52 lines
2.9 KiB
Markdown
# Grafana dashboard
|
|
|
|
`rustproxy-overview.json` is a provisioning/import dashboard for the dedicated
|
|
RustProxy Prometheus exporter. It covers global traffic, status classes, vhost
|
|
latency, backend health/latency, upstream-client generation/retry/rotation diagnostics,
|
|
a dedicated Docker-vhost section, hot/cold/transient cache memory, Linux process and cgroup diagnostics, HTTP/2/HTTP/3/TLS/WebSocket
|
|
counters, Layer-4 streams, structured-log sink health, configuration inventory and bounded Traffic analytics.
|
|
|
|
The **Docker vhosts** row uses only the dedicated `rust_proxy_docker_*` metric
|
|
families and includes configured/enabled counts, request/error rates, p95 exchange
|
|
latency, backend health/EWMA latency/active work, traffic, probe age and upstream-client lifecycle/retry panels. Dashboard
|
|
variables **Docker agent** and **Docker service** filter only this section, which keeps
|
|
multi-agent installations readable without changing the ordinary vhost panels.
|
|
|
|
The **Web Server** row adds header-ready request rates, status-class and
|
|
static/PHP/other donut breakdowns, error percentage, response-header latency,
|
|
payload byte rates, and downstream compression by codec. **PHP-FPM / FastCGI**
|
|
adds request activity, RustProxy PHP permit capacity, and errors grouped by
|
|
finite reason. **Per Web Service** shows top service rates, HTTP methods and
|
|
selected status codes, while **WWW / PHP correlation** compares PHP load and
|
|
Web Server errors with cache activity. Use the **Web service** variable to
|
|
filter the new panels by configured Web Service ID. All new panels are
|
|
added to the existing dashboard without changing the legacy rows.
|
|
|
|
The **Traffic analytics and Web Server** row uses the `rust_proxy_traffic_*` families.
|
|
The **Traffic vhost** variable filters that row and exposes Web Server request rate,
|
|
serving-source rate, top countries, fixed User-Agent families, HTTP methods, status
|
|
classes, cache outcomes, capture-state gauges and configured-route traffic. Raw IP addresses, request paths and full User-Agent strings
|
|
are intentionally absent from Prometheus labels.
|
|
|
|
Import the JSON in Grafana and select a Prometheus datasource when prompted. The
|
|
dashboard expects a RustProxy version that exports the Web Server metric names
|
|
(and also retains the older panels); a scrape interval around 15s is suitable.
|
|
Collectors can be disabled in RustProxy settings; panels for disabled collectors
|
|
will simply show no series.
|
|
|
|
Recommended scrape job:
|
|
|
|
```yaml
|
|
scrape_configs:
|
|
- job_name: rustproxy
|
|
scrape_interval: 15s
|
|
metrics_path: /metrics
|
|
static_configs:
|
|
- targets: ['rustproxy-host:9191']
|
|
```
|
|
|
|
When the exporter is protected with a bearer token, configure Prometheus
|
|
`authorization.credentials_file`. For a private monitoring VLAN, RustProxy can
|
|
instead explicitly allow actual peers from configured trusted CIDRs to scrape
|
|
without a token. Forwarded client-IP headers never satisfy that trust check.
|