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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
//! Process spawning and health check execution.
use crate::sdk::{FailureReason, ServiceConfig, ServiceState};
use crate::server::graph::ServiceId;
use crate::server::log;
use crate::server::process::{self, SpawnResult, check_health};
use super::Supervisor;
use super::events::{StartAction, SupervisorEvent, TimeoutKind};
impl Supervisor {
/// Try to start a service if possible.
pub(crate) async fn try_start_service(&mut self, id: ServiceId) {
// Collect all info needed under lock, then release it
let action = {
let mut graph = self.graph.write().await;
let service = match graph.get(id) {
Some(s) => s,
None => return,
};
// Check desired status - if "stop" or "ignore", don't auto-start
if !service.should_autostart() {
tracing::trace!(service = %service.name, "skipping start: status != start");
return;
}
// Check if we can attempt start
if !service.state.can_attempt_start() {
tracing::debug!(
service = %service.name,
state = %service.state,
"skipping start: state does not allow start attempts"
);
return;
}
// Check dependencies
match graph.can_start(id) {
Ok(()) => {
// Can start!
if service.is_target() {
// Targets go directly to Running
let name = service.name.clone();
graph.set_state(id, ServiceState::Running { pid: 0 });
let dependents = graph.dependents(id);
StartAction::TargetReady { name, dependents }
} else {
// Real service needs to be spawned
let config = service.service_config().cloned();
graph.set_state(id, ServiceState::Starting { pid: 0 });
StartAction::SpawnProcess { config }
}
}
Err(reason) => {
// Blocked
let name = service.name.clone();
graph.set_state(
id,
ServiceState::Blocked {
waiting_on: reason.waiting_on(),
},
);
StartAction::Blocked {
name,
reason: reason.to_string(),
}
}
}
};
// Lock released here
// Now perform actions without holding the lock
match action {
StartAction::TargetReady { name, dependents } => {
tracing::info!(service = %name, "target satisfied");
self.queue_reevaluate(dependents).await;
}
StartAction::SpawnProcess { config } => {
if let Some(config) = config {
// Check for builtin service
if let Some(builtin) = Self::is_builtin_service(&config) {
self.run_builtin_service(id, builtin, &config).await;
} else {
self.spawn_service(id, config).await;
}
}
}
StartAction::Blocked { name, reason } => {
tracing::debug!(service = %name, reason = %reason, "service blocked");
}
}
}
/// Queue services for re-evaluation.
pub(crate) async fn queue_reevaluate(&self, service_ids: Vec<ServiceId>) {
for id in service_ids {
let _ = self
.event_tx
.send(SupervisorEvent::Reevaluate { service_id: id })
.await;
}
}
/// Spawn a service process.
pub(crate) async fn spawn_service(&mut self, id: ServiceId, config: ServiceConfig) {
let name = config.service.name.clone();
// Initialize log buffer
let buffer_lines = config.logging.buffer_lines;
log::init_buffer(&self.log_buffers, id, buffer_lines).await;
// Kill other processes on declared ports if kill_others is enabled
if config.service.kill_others && !config.service.ports.is_empty() {
tracing::info!(
service = %name,
ports = ?config.service.ports,
"kill_others enabled, killing processes on these ports"
);
let kill_result = process::kill_processes_on_ports(&config.service.ports);
if !kill_result.success {
tracing::warn!(
service = %name,
failed_pids = ?kill_result.failed,
"some processes could not be killed, proceeding anyway"
);
} else if !kill_result.killed.is_empty() {
tracing::info!(
service = %name,
killed_count = kill_result.killed.len(),
killed_pids = ?kill_result.killed,
"successfully killed processes on ports"
);
}
}
// Kill other processes matching any filter if kill_others is enabled
if config.service.kill_others && !config.service.process_filters.is_empty() {
tracing::info!(
service = %name,
filters = ?config.service.process_filters,
"kill_others enabled, killing processes matching filters"
);
let mut all_processes = Vec::new();
for filter in &config.service.process_filters {
let processes = crate::server::graph::find_processes_by_name(filter);
for proc_info in processes {
all_processes.push((filter.clone(), proc_info));
}
}
if !all_processes.is_empty() {
tracing::info!(
service = %name,
count = all_processes.len(),
pids = ?all_processes.iter().map(|(_, p)| p.pid).collect::<Vec<_>>(),
"killing processes matching filters"
);
for (_filter, proc_info) in all_processes {
let kill_result = process::kill_process_tree(proc_info.pid);
if !kill_result.success {
tracing::warn!(
service = %name,
pid = proc_info.pid,
process_name = %proc_info.name,
failed_pids = ?kill_result.failed,
"failed to kill some processes in tree"
);
} else if !kill_result.killed.is_empty() {
tracing::info!(
service = %name,
pid = proc_info.pid,
process_name = %proc_info.name,
killed_count = kill_result.killed.len(),
killed_pids = ?kill_result.killed,
"successfully killed process tree"
);
}
}
}
}
// Spawn the process
match process::spawn_process(&config) {
Ok(SpawnResult {
child,
stdout,
stderr,
pid,
}) => {
// Update state with actual PID and record start time
{
let mut graph = self.graph.write().await;
if let Some(service) = graph.get_mut(id) {
service.record_started();
service.state = ServiceState::Starting { pid };
}
}
// Start log readers
let log_buffers = self.log_buffers.clone();
let name_clone = name.clone();
tokio::spawn(async move {
log::read_stdout(id, name_clone, stdout, log_buffers, None).await;
});
let log_buffers = self.log_buffers.clone();
let name_clone = name.clone();
tokio::spawn(async move {
log::read_stderr(id, name_clone, stderr, log_buffers, None).await;
});
// Start process watcher
let event_tx = self.event_tx.clone();
let task = tokio::spawn(async move {
let (exit_code, signal) = process::wait_for_exit(child).await;
let _ = event_tx
.send(SupervisorEvent::ProcessExited {
service_id: id,
exit_code,
signal,
})
.await;
});
self.process_tasks.insert(id, task);
// Schedule default start timeout (may be extended for health checks)
self.schedule_timeout(id, TimeoutKind::Start, config.lifecycle.start_timeout_ms);
// Check if service has health checks
if let Some(ref health) = config.health {
// Extract health check timing parameters
let (start_period, interval, retries, timeout) = match health {
crate::sdk::HealthDef::Tcp { common, .. }
| crate::sdk::HealthDef::Http { common, .. }
| crate::sdk::HealthDef::Exec { common, .. } => (
common.start_period_ms,
common.interval_ms,
common.retries,
common.timeout_ms,
),
};
// Auto-extend start_timeout to accommodate health checks
// Total time needed: start_period + (retries * (interval + timeout)) + buffer
let health_time_needed =
start_period + (retries as u64 * (interval + timeout)) + 5000;
let effective_start_timeout =
config.lifecycle.start_timeout_ms.max(health_time_needed);
if effective_start_timeout > config.lifecycle.start_timeout_ms {
tracing::debug!(
service = %name,
configured = config.lifecycle.start_timeout_ms,
effective = effective_start_timeout,
"auto-extended start_timeout to accommodate health checks"
);
}
self.schedule_timeout(id, TimeoutKind::Start, effective_start_timeout);
if start_period > 0 {
self.schedule_timeout(id, TimeoutKind::HealthCheck, start_period);
} else {
// Run health check immediately
self.run_health_check(id, health.clone());
}
} else {
// No health check - transition directly to Running
let dependents = {
let mut graph = self.graph.write().await;
if let Some(service) = graph.get_mut(id) {
service.state = ServiceState::Running { pid };
// Note: Don't reset backoff here - it will be reset when the
// service has been running long enough (stability period).
// This happens when the service exits/stops via try_reset_backoff().
}
graph.dependents(id)
};
self.cancel_timeout(id, TimeoutKind::Start);
self.queue_reevaluate(dependents).await;
}
tracing::info!(service = %name, pid = pid, "service started");
}
Err(e) => {
tracing::error!(service = %name, error = %e, "failed to spawn service");
let mut graph = self.graph.write().await;
graph.set_state(
id,
ServiceState::Failed {
reason: FailureReason::SpawnError {
message: e.to_string(),
},
},
);
}
}
}
/// Run a health check for a service.
pub(crate) fn run_health_check(&self, id: ServiceId, health: crate::sdk::HealthDef) {
let event_tx = self.event_tx.clone();
tokio::spawn(async move {
let result = check_health(&health).await;
let _ = event_tx
.send(SupervisorEvent::HealthCheckResult {
service_id: id,
passed: result.is_ok(),
error: result.err().map(|e| e.to_string()),
})
.await;
});
}
}