File size: 15,142 Bytes
d02897f | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 | """
OpenOps Production Incident Management Environment
Simulates real-world production incidents where agents must investigate, mitigate, and resolve issues
"""
from typing import Dict, List, Any, Optional
from openenv.core import Environment
from models import IncidentAction, IncidentObservation, IncidentState
class MyEnvEnvironment(Environment):
"""
Production incident management environment.
Simulates 3 types of incidents:
- Task 1 (Easy): Simple API crash requiring restart
- Task 2 (Medium): Bad deployment requiring rollback
- Task 3 (Hard): Cascading database overload requiring multi-step resolution
"""
# Action definitions
ACTION_NAMES = {
# Investigation actions (0-8)
0: "read_alerts",
1: "inspect_logs_api",
2: "inspect_logs_database",
3: "inspect_logs_auth",
4: "inspect_logs_frontend",
5: "check_metrics_api",
6: "check_metrics_database",
7: "check_metrics_auth",
8: "check_metrics_frontend",
# Mitigation actions (9-16)
9: "restart_api",
10: "restart_database",
11: "restart_auth",
12: "restart_frontend",
13: "rollback_api",
14: "rollback_database",
15: "scale_api",
16: "scale_database",
# Communication actions (17-19)
17: "notify_team",
18: "update_status_page",
19: "send_user_communication",
# Resolution (20)
20: "resolve_incident"
}
def __init__(self):
"""Initialize the environment."""
super().__init__()
self.task_id = 1
self.time_elapsed = 0
self.max_steps = 30
self.total_reward = 0.0
# State tracking
self.incident_resolved = False
self.alerts_read = False
self.logs_inspected = set()
self.metrics_checked = set()
self.services_restarted = set()
self.services_rolled_back = set()
self.services_scaled = set()
self.teams_notified = False
self.status_page_updated = False
self.users_communicated = False
# Internal state
self._state = None
@property
def state(self) -> IncidentState:
"""
Return current environment state.
Required by BaseEnvironment abstract class.
"""
return self._state
@state.setter
def state(self, value: IncidentState):
"""Set the environment state."""
self._state = value
def reset(self, task_id: int = 1) -> IncidentObservation:
"""
Reset environment for a specific task.
Args:
task_id: Task difficulty (1=easy, 2=medium, 3=hard)
Returns:
Initial observation
"""
self.task_id = task_id
self.time_elapsed = 0
self.total_reward = 0.0
# Reset tracking
self.incident_resolved = False
self.alerts_read = False
self.logs_inspected = set()
self.metrics_checked = set()
self.services_restarted = set()
self.services_rolled_back = set()
self.services_scaled = set()
self.teams_notified = False
self.status_page_updated = False
self.users_communicated = False
# Initialize state based on task
self._state = self._init_task_state(task_id)
# Return initial observation
return self._get_observation()
def _init_task_state(self, task_id: int) -> IncidentState:
"""Initialize task-specific state."""
if task_id == 1:
# Task 1: Simple API crash (OOM)
return IncidentState(
task_id=task_id,
incident_type="api_crash",
affected_services=["api"],
root_cause="out_of_memory",
service_status={
"api": "down",
"database": "healthy",
"auth": "healthy",
"frontend": "degraded"
},
correct_mitigation=["restart_api"],
revenue_loss=0.0,
customer_complaints=0
)
elif task_id == 2:
# Task 2: Bad deployment (database)
return IncidentState(
task_id=task_id,
incident_type="bad_deployment",
affected_services=["database", "api"],
root_cause="bad_migration",
service_status={
"api": "degraded",
"database": "degraded",
"auth": "healthy",
"frontend": "degraded"
},
correct_mitigation=["rollback_database"],
revenue_loss=0.0,
customer_complaints=0
)
else: # task_id == 3
# Task 3: Cascading failure (database overload)
return IncidentState(
task_id=task_id,
incident_type="cascading_failure",
affected_services=["database", "api"],
root_cause="database_overload",
service_status={
"api": "degraded",
"database": "degraded",
"auth": "healthy",
"frontend": "degraded"
},
correct_mitigation=["scale_database", "restart_api"],
revenue_loss=0.0,
customer_complaints=0
)
def step(self, action: IncidentAction) -> IncidentObservation:
"""
Execute an action and return observation.
Args:
action: Action to execute
Returns:
Observation after action execution
"""
self.time_elapsed += 1
reward = 0.0
done = False
# Time penalty
reward -= 0.05
# Revenue loss increases over time
self._state.revenue_loss += 1000 * self.time_elapsed
self._state.customer_complaints += self.time_elapsed // 3
# Execute action
action_name = self.ACTION_NAMES.get(action.action_id, "unknown")
# Investigation actions
if action.action_id == 0: # read_alerts
if not self.alerts_read:
self.alerts_read = True
reward += 0.05
elif 1 <= action.action_id <= 4: # inspect_logs
service = ["api", "database", "auth", "frontend"][action.action_id - 1]
if service not in self.logs_inspected:
self.logs_inspected.add(service)
if service in self._state.affected_services:
reward += 0.25 # Bonus for inspecting affected service
else:
reward += 0.05
elif 5 <= action.action_id <= 8: # check_metrics
service = ["api", "database", "auth", "frontend"][action.action_id - 5]
if service not in self.metrics_checked:
self.metrics_checked.add(service)
reward += 0.05
# Mitigation actions
elif 9 <= action.action_id <= 12: # restart services
service = ["api", "database", "auth", "frontend"][action.action_id - 9]
if service not in self.services_restarted:
self.services_restarted.add(service)
# Check if restart is correct mitigation
if "restart_" + service in self._state.correct_mitigation:
reward += 0.75
self._state.service_status[service] = "healthy"
elif service in self._state.affected_services:
# Restarting affected service (but not the solution)
reward -= 0.5
else:
reward -= 0.1
elif 13 <= action.action_id <= 14: # rollback (API or Database)
service = ["api", "database"][action.action_id - 13]
if service not in self.services_rolled_back:
self.services_rolled_back.add(service)
# Check if rollback is correct
if "rollback_" + service in self._state.correct_mitigation:
reward += 1.0
self._state.service_status[service] = "healthy"
if service == "database":
self._state.service_status["api"] = "healthy" # Fixes downstream
else:
reward -= 0.3
elif 15 <= action.action_id <= 16: # scale (API or Database)
service = ["api", "database"][action.action_id - 15]
if service not in self.services_scaled:
self.services_scaled.add(service)
# Check if scaling is correct
if "scale_" + service in self._state.correct_mitigation:
reward += 0.75
self._state.service_status[service] = "healthy"
else:
reward -= 0.2
# Communication actions
elif action.action_id == 17: # notify_team
if not self.teams_notified:
self.teams_notified = True
reward += 0.25
elif action.action_id == 18: # update_status_page
if not self.status_page_updated:
self.status_page_updated = True
reward += 0.25
elif action.action_id == 19: # send_user_communication
if not self.users_communicated:
self.users_communicated = True
reward += 0.15
# Resolution
elif action.action_id == 20: # resolve_incident
# Check if all services are healthy
all_healthy = all(
status == "healthy"
for service, status in self._state.service_status.items()
if service in self._state.affected_services
)
if all_healthy:
self.incident_resolved = True
# Big reward for resolution
reward += 3.0
# Time bonus (faster = better)
time_bonus = max(0, (30 - self.time_elapsed) * 0.01)
reward += time_bonus
done = True
else:
# Penalty for premature resolution
reward -= 1.0
done = True
# Update total reward
self.total_reward += reward
# Check timeout
if self.time_elapsed >= self.max_steps:
done = True
# Return observation
obs = self._get_observation()
obs.reward = reward
obs.done = done
return obs
def _get_observation(self) -> IncidentObservation:
"""Generate current observation."""
# Build alerts
active_alerts = []
if not self.alerts_read:
active_alerts = ["[Call read_alerts to see alerts]"]
else:
if self._state.task_id == 1:
active_alerts = [
"🚨 CRITICAL: API service down - no response",
"⚠️ HIGH: Frontend experiencing errors",
"📊 Customer complaints spiking"
]
elif self._state.task_id == 2:
active_alerts = [
"🚨 CRITICAL: Database queries failing",
"⚠️ HIGH: API returning 500 errors",
"📊 Recent deployment detected"
]
else: # task_id == 3
active_alerts = [
"🚨 CRITICAL: Database CPU at 95%",
"🚨 CRITICAL: API timeout rate high",
"⚠️ HIGH: Connection pool exhausted",
"📊 Cascading failure detected"
]
# Build logs (only for inspected services)
recent_logs = {}
for service in self.logs_inspected:
if self._state.task_id == 1 and service == "api":
recent_logs["api"] = [
"ERROR: Out of memory - process killed",
"INFO: Last request before crash at 14:32:15"
]
elif self._state.task_id == 2 and service == "database":
recent_logs["database"] = [
"ERROR: Syntax error in migration v2.3.1",
"ERROR: Incompatible schema changes detected"
]
elif self._state.task_id == 2 and service == "api":
recent_logs["api"] = [
"ERROR: Database query timeout",
"ERROR: 500 Internal Server Error"
]
elif self._state.task_id == 3 and service == "database":
recent_logs["database"] = [
"WARN: Connection pool exhausted (95% utilization)",
"ERROR: Slow query detected (>10s)",
"WARN: CPU usage at 95%"
]
elif self._state.task_id == 3 and service == "api":
recent_logs["api"] = [
"ERROR: Database connection timeout",
"ERROR: Request timeout (30s exceeded)"
]
# Build metrics summary
metrics_summary = {}
for service in self.metrics_checked:
if service in self._state.affected_services:
metrics_summary[service] = {
"cpu": 85.0 if service == "database" else 45.0,
"memory": 92.0 if service == "api" else 60.0,
"latency": 5000.0 if service in ["api", "database"] else 100.0
}
return IncidentObservation(
active_alerts=active_alerts,
service_status=self._state.service_status.copy(),
recent_logs=recent_logs,
metrics_summary=metrics_summary,
customer_complaints=self._state.customer_complaints,
time_elapsed=self.time_elapsed,
revenue_loss=self._state.revenue_loss,
teams_notified=self.teams_notified,
status_page_updated=self.status_page_updated,
reward=0.0,
done=False
)
def render(self):
"""Render current state (optional for debugging)."""
print(f"\n{'='*60}")
print(f"Task {self.task_id} - Step {self.time_elapsed}")
print(f"{'='*60}")
print(f"Service Status: {self._state.service_status}")
print(f"Revenue Loss: ${self._state.revenue_loss:,.0f}")
print(f"Complaints: {self._state.customer_complaints}")
print(f"Incident Resolved: {self.incident_resolved}")
print(f"Total Reward: {self.total_reward:.2f}")
print(f"{'='*60}\n") |