referrerpolicy=no-referrer-when-downgrade

relay_utils/metrics/
global.rs

1// Copyright 2019-2021 Parity Technologies (UK) Ltd.
2// This file is part of Parity Bridges Common.
3
4// Parity Bridges Common is free software: you can redistribute it and/or modify
5// it under the terms of the GNU General Public License as published by
6// the Free Software Foundation, either version 3 of the License, or
7// (at your option) any later version.
8
9// Parity Bridges Common is distributed in the hope that it will be useful,
10// but WITHOUT ANY WARRANTY; without even the implied warranty of
11// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12// GNU General Public License for more details.
13
14// You should have received a copy of the GNU General Public License
15// along with Parity Bridges Common.  If not, see <http://www.gnu.org/licenses/>.
16
17//! Global system-wide Prometheus metrics exposed by relays.
18
19use crate::metrics::{
20	metric_name, register, Gauge, GaugeVec, Metric, Opts, PrometheusError, Registry,
21	StandaloneMetric, F64, U64,
22};
23
24use async_std::sync::{Arc, Mutex};
25use async_trait::async_trait;
26use std::time::Duration;
27use sysinfo::{RefreshKind, System};
28
29/// Global metrics update interval.
30const UPDATE_INTERVAL: Duration = Duration::from_secs(10);
31
32/// Global Prometheus metrics.
33#[derive(Debug, Clone)]
34pub struct GlobalMetrics {
35	system: Arc<Mutex<System>>,
36	system_average_load: GaugeVec<F64>,
37	process_cpu_usage_percentage: Gauge<F64>,
38	process_memory_usage_bytes: Gauge<U64>,
39}
40
41impl GlobalMetrics {
42	/// Create and register global metrics.
43	pub fn new() -> Result<Self, PrometheusError> {
44		Ok(GlobalMetrics {
45			system: Arc::new(Mutex::new(System::new_with_specifics(RefreshKind::everything()))),
46			system_average_load: GaugeVec::new(
47				Opts::new(metric_name(None, "system_average_load"), "System load average"),
48				&["over"],
49			)?,
50			process_cpu_usage_percentage: Gauge::new(
51				metric_name(None, "process_cpu_usage_percentage"),
52				"Process CPU usage",
53			)?,
54			process_memory_usage_bytes: Gauge::new(
55				metric_name(None, "process_memory_usage_bytes"),
56				"Process memory (resident set size) usage",
57			)?,
58		})
59	}
60}
61
62impl Metric for GlobalMetrics {
63	fn register(&self, registry: &Registry) -> Result<(), PrometheusError> {
64		register(self.system_average_load.clone(), registry)?;
65		register(self.process_cpu_usage_percentage.clone(), registry)?;
66		register(self.process_memory_usage_bytes.clone(), registry)?;
67		Ok(())
68	}
69}
70
71#[async_trait]
72impl StandaloneMetric for GlobalMetrics {
73	async fn update(&self) {
74		// update system-wide metrics
75		let mut system = self.system.lock().await;
76		let load = sysinfo::System::load_average();
77		self.system_average_load.with_label_values(&["1min"]).set(load.one);
78		self.system_average_load.with_label_values(&["5min"]).set(load.five);
79		self.system_average_load.with_label_values(&["15min"]).set(load.fifteen);
80
81		// update process-related metrics
82		let pid = sysinfo::get_current_pid().expect(
83			"only fails where pid is unavailable (os=unknown || arch=wasm32);\
84				relay is not supposed to run in such MetricsParamss;\
85				qed",
86		);
87		let is_process_refreshed = system.refresh_process(pid);
88		match (is_process_refreshed, system.process(pid)) {
89			(true, Some(process_info)) => {
90				let cpu_usage = process_info.cpu_usage() as f64;
91				let memory_usage = process_info.memory() * 1024;
92				log::trace!(
93					target: "bridge-metrics",
94					"Refreshed process metrics: CPU={}, memory={}",
95					cpu_usage,
96					memory_usage,
97				);
98
99				self.process_cpu_usage_percentage.set(if cpu_usage.is_finite() {
100					cpu_usage
101				} else {
102					0f64
103				});
104				self.process_memory_usage_bytes.set(memory_usage);
105			},
106			_ => {
107				log::warn!(
108					target: "bridge-metrics",
109					"Failed to refresh process information. Metrics may show obsolete values",
110				);
111			},
112		}
113	}
114
115	fn update_interval(&self) -> Duration {
116		UPDATE_INTERVAL
117	}
118}