1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
|
use std::mem::MaybeUninit;
use bollard::container::{BlkioStats, BlkioStatsEntry, StatsOptions};
use bollard::models::ContainerSummary;
use bollard::Docker;
use opentelemetry::metrics::MeterProvider;
use opentelemetry::KeyValue;
use opentelemetry_sdk::metrics::SdkMeterProvider;
use std::sync::Arc;
use std::time::{Duration, SystemTime, UNIX_EPOCH};
use tokio::task::JoinHandle;
use tokio_stream::StreamExt;
// I do not enjoy taking a bunch of Rcs but tokio needs ownership so fine.
pub fn launch_stats_task(
container: ContainerSummary,
docker: Arc<Docker>,
meter_provider: Arc<SdkMeterProvider>,
) -> JoinHandle<()> {
tokio::spawn(async move {
// extract some container info
let container_id = container.id.unwrap();
let container_name = container
.names
.iter()
.flatten()
.next()
.map(|n| n.trim_start_matches("/").to_owned());
let meter_name = "cspy_".to_string() + container_id.as_str();
// lol 'static moment
let meter_name = &*Box::leak(meter_name.into_boxed_str());
let mut stats_stream = docker.stats(
container_id.as_str(),
Some(StatsOptions {
stream: true,
one_shot: false,
}),
);
// use the first read only for stats diffing for blkio - don't need for cpu thanks to precpu.
#[allow(unused_assignments)]
let mut first_read = MaybeUninit::uninit();
loop {
match stats_stream.next().await {
None => return,
Some(Ok(st)) => { first_read = MaybeUninit::new(st); break },
Some(Err(err)) => {
// TODO: use json logging or syslog so loki can understand this lol
println!("Failed to get stats for container {container_id}!: {err:?}");
}
}
}
// I'm going to rust jail!
let first_read = unsafe { first_read.assume_init() };
let mut last_io_stats = first_read.blkio_stats.io_service_bytes_recursive;
// container labels shared for all metrics
let mut shared_labels = vec![
KeyValue::new("id", container_id.to_owned()),
KeyValue::new(
"image",
container.image.unwrap_or(container.image_id.unwrap()),
),
];
if let Some(name) = container_name {
shared_labels.push(KeyValue::new("name", name));
}
if let Some(docker_labels) = &container.labels {
for (key, value) in docker_labels {
shared_labels.push(KeyValue::new(
"container_label_".to_string() + key,
value.clone(),
))
}
}
// free space and make mutable
shared_labels.shrink_to_fit();
let shared_labels = &shared_labels[..];
//println!("Starting reporting for container: {shared_labels:?}");
// create meters
let meter = meter_provider.meter(meter_name);
let meter_container_cpu_usage_seconds_total = meter
.f64_counter("container_cpu_usage_seconds_total")
.with_unit("s")
.with_description("Cumulative cpu time consumed")
.build();
let meter_container_cpu_user_seconds_total = meter
.f64_counter("container_cpu_user_seconds_total")
.with_unit("s")
.with_description("Cumulative userland cpu time consumed")
.build();
let meter_container_cpu_system_seconds_total = meter
.f64_counter("container_cpu_system_seconds_total")
.with_unit("s")
.with_description("Cumulative kernel cpu time consumed")
.build();
let meter_container_cpu_cfs_periods_total = meter
.u64_counter("container_cpu_cfs_periods_total")
.with_description("Number of elapsed enforcement period intervals")
.build();
let meter_container_cpu_cfs_throttled_periods_total = meter
.u64_counter("container_cpu_cfs_throttled_periods_total")
.with_description("Number of throttled period intervals")
.build();
let meter_container_cpu_cfs_throttled_seconds_total = meter
.f64_counter("container_cpu_cfs_throttled_seconds_total")
.with_unit("s")
.with_description("Total time duration the container has been throttled")
.build();
let meter_container_fs_reads_bytes_total = meter
.u64_counter("container_fs_reads_bytes_total")
.with_unit("By")
.with_description("Cumulative bytes read")
.build();
let meter_container_fs_writes_bytes_total = meter
.u64_counter("container_fs_writes_bytes_total")
.with_unit("By")
.with_description("Cumulative bytes written")
.build();
let meter_container_last_seen = meter
.u64_gauge("container_last_seen")
.with_description("Last time this container was seen by ContainerSpy")
.build();
while let Some(val) = stats_stream.next().await {
if let Ok(stats) = val {
meter_container_cpu_usage_seconds_total.add(
cpu_delta_from_docker(
stats.cpu_stats.cpu_usage.total_usage,
stats.precpu_stats.cpu_usage.total_usage,
)
.as_secs_f64(),
shared_labels,
);
meter_container_cpu_user_seconds_total.add(
cpu_delta_from_docker(
stats.cpu_stats.cpu_usage.usage_in_usermode,
stats.precpu_stats.cpu_usage.usage_in_usermode,
)
.as_secs_f64(),
shared_labels,
);
meter_container_cpu_system_seconds_total.add(
cpu_delta_from_docker(
stats.cpu_stats.cpu_usage.usage_in_kernelmode,
stats.precpu_stats.cpu_usage.usage_in_kernelmode,
)
.as_secs_f64(),
shared_labels,
);
meter_container_cpu_cfs_periods_total.add(
stats.cpu_stats.throttling_data.periods - stats.precpu_stats.throttling_data.periods,
shared_labels,
);
meter_container_cpu_cfs_throttled_periods_total.add(
stats.cpu_stats.throttling_data.throttled_periods
- stats.precpu_stats.throttling_data.throttled_periods,
shared_labels,
);
meter_container_cpu_cfs_throttled_seconds_total.add(
cpu_delta_from_docker(stats.cpu_stats.throttling_data.throttled_time,
stats.precpu_stats.throttling_data.throttled_time).as_secs_f64(),
shared_labels,
);
// other blkio_stats values only exist on cgroups v1 so don't bother.
// io_service_bytes_recursive exists only on cgroups v1.
// storage_stats only exists on windows.
if let Some(service_bytes_rec) = stats.blkio_stats.io_service_bytes_recursive {
// need to calculate deltas for this
if let Some(last) = &last_io_stats {
let (last_r, last_w) = get_rw_totals(last);
let (curr_r, curr_w) = get_rw_totals(&service_bytes_rec);
meter_container_fs_reads_bytes_total.add(curr_r - last_r, shared_labels);
meter_container_fs_writes_bytes_total.add(curr_w - last_w, shared_labels);
}
last_io_stats = Some(service_bytes_rec);
}
meter_container_last_seen.record(SystemTime::now().duration_since(UNIX_EPOCH).unwrap().as_secs(), shared_labels);
} else {
// failed to get stats, log as such:
// TODO: use json logging or syslog so loki can understand this lol
println!(
"Failed to get stats for container {container_id}!: {:?}",
val.unwrap_err()
);
}
}
})
}
fn cpu_delta_from_docker(cpu_usage: u64, precpu_usage: u64) -> Duration {
let delta = cpu_usage - precpu_usage;
// https://docs.docker.com/reference/api/engine/version/v1.48/#tag/Container/operation/ContainerStats
// see response schema > cpu_stats > cpu_usage > total_usage
let delta_ns = if cfg!(windows) { delta * 100 } else { delta };
Duration::from_nanos(delta_ns)
}
fn get_rw_totals<'a>(iter: impl IntoIterator<Item = &'a BlkioStatsEntry>) -> (u64, u64) {
let mut read = 0;
let mut write = 0;
for entry in iter {
match entry.op.as_str() {
"read" => read += entry.value,
"write" => write += entry.value,
_ => println!("Unknown service_bytes_recursive entry type {}", entry.op)
}
}
(read, write)
}
|