overgrowth / agent /deployment_engine.py
Graham Paasch
test: Stage 6 comprehensive test suite - 23/23 passing
d36860e
Raw
History Blame
18.6 kB
"""
Deployment orchestration engine.
Orchestrates end-to-end configuration deployment:
- Config generation from templates
- Pre-deployment validation
- Device connectivity and deployment
- Post-deployment verification
- Automatic rollback on failure
- Parallel deployment via Ray integration
"""
import logging
from typing import Dict, List, Optional, Any, Tuple
from dataclasses import dataclass, field
from enum import Enum
from datetime import datetime
import time
from agent.device_driver import (
DeviceDriver, DeviceCredentials, DeviceType,
ConfigDeploymentResult, ConnectionStatus
)
from agent.config_templates import ConfigTemplateEngine
logger = logging.getLogger(__name__)
class DeploymentStatus(Enum):
"""Deployment status"""
PENDING = "pending"
VALIDATING = "validating"
DEPLOYING = "deploying"
VERIFYING = "verifying"
SUCCESS = "success"
FAILED = "failed"
ROLLED_BACK = "rolled_back"
@dataclass
class DeploymentTask:
"""Single device deployment task"""
device_id: str
device_type: DeviceType
hostname: str
username: str
password: str
config: str
dry_run: bool = False
pre_checks: List[str] = field(default_factory=list)
post_checks: List[str] = field(default_factory=list)
@dataclass
class DeploymentResult:
"""Result from deployment orchestration"""
device_id: str
status: DeploymentStatus
config_deployed: Optional[str] = None
config_before: Optional[str] = None
config_after: Optional[str] = None
pre_check_results: Dict[str, bool] = field(default_factory=dict)
post_check_results: Dict[str, bool] = field(default_factory=dict)
deployment_output: Optional[str] = None
error: Optional[str] = None
duration_seconds: float = 0.0
rolled_back: bool = False
timestamp: datetime = field(default_factory=datetime.now)
class DeploymentEngine:
"""
Orchestrates configuration deployment to network devices.
Features:
- Multi-vendor device support (Cisco, Arista, Juniper)
- Pre/post deployment validation
- Automatic rollback on failure
- Parallel deployment via Ray
- Dry-run mode for testing
"""
def __init__(self, use_napalm: bool = True, use_ray: bool = False):
"""
Initialize deployment engine.
Args:
use_napalm: Prefer NAPALM over Netmiko
use_ray: Use Ray for parallel deployment
"""
self.driver = DeviceDriver(use_napalm=use_napalm)
self.template_engine = ConfigTemplateEngine()
self.use_ray = use_ray
if use_ray:
try:
from agent.ray_executor import RayExecutor
self.ray_executor = RayExecutor()
logger.info("DeploymentEngine using Ray for parallel execution")
except ImportError:
logger.warning("Ray not available - falling back to serial execution")
self.use_ray = False
self.ray_executor = None
else:
self.ray_executor = None
self.deployment_history: List[DeploymentResult] = []
def deploy_single_device(self, task: DeploymentTask) -> DeploymentResult:
"""
Deploy configuration to a single device.
Args:
task: Deployment task specification
Returns:
DeploymentResult with outcome
"""
logger.info(f"Starting deployment to {task.device_id} (dry_run={task.dry_run})")
start_time = time.time()
result = DeploymentResult(
device_id=task.device_id,
status=DeploymentStatus.PENDING
)
try:
# 1. Connect to device
logger.info(f"Connecting to {task.device_id}...")
credentials = DeviceCredentials(
hostname=task.hostname,
username=task.username,
password=task.password,
device_type=task.device_type
)
conn = self.driver.connect(credentials)
if conn.status != ConnectionStatus.CONNECTED:
raise Exception(f"Failed to connect: {conn.last_error}")
# 2. Pre-deployment checks
result.status = DeploymentStatus.VALIDATING
logger.info(f"Running {len(task.pre_checks)} pre-deployment checks...")
for check in task.pre_checks:
check_result = self._run_validation_check(task.device_id, check)
result.pre_check_results[check] = check_result
if not check_result:
raise Exception(f"Pre-deployment check failed: {check}")
# 3. Get current config (for rollback)
result.config_before = self.driver.get_config(task.device_id, "running")
# 4. Deploy configuration
result.status = DeploymentStatus.DEPLOYING
logger.info(f"Deploying configuration to {task.device_id}...")
deploy_result = self.driver.deploy_config(
device_id=task.device_id,
config=task.config,
dry_run=task.dry_run,
replace=False # Merge by default
)
if not deploy_result.success:
raise Exception(f"Deployment failed: {deploy_result.error}")
result.config_deployed = task.config
result.config_after = deploy_result.config_after
result.deployment_output = deploy_result.output
# 5. Post-deployment verification
if not task.dry_run:
result.status = DeploymentStatus.VERIFYING
logger.info(f"Running {len(task.post_checks)} post-deployment checks...")
# Give device time to apply config
time.sleep(2)
for check in task.post_checks:
check_result = self._run_validation_check(task.device_id, check)
result.post_check_results[check] = check_result
if not check_result:
# Post-check failed - rollback!
logger.error(f"Post-deployment check failed: {check}")
logger.warning(f"Initiating rollback on {task.device_id}")
rollback_result = self.driver.rollback_config(
device_id=task.device_id,
config=result.config_before
)
result.rolled_back = rollback_result.success
result.status = DeploymentStatus.ROLLED_BACK if result.rolled_back else DeploymentStatus.FAILED
raise Exception(f"Post-deployment check failed: {check} (rollback {'succeeded' if result.rolled_back else 'FAILED'})")
# Success!
result.status = DeploymentStatus.SUCCESS
result.duration_seconds = time.time() - start_time
logger.info(f"✓ Deployment to {task.device_id} completed successfully in {result.duration_seconds:.1f}s")
except Exception as e:
result.status = DeploymentStatus.FAILED
result.error = str(e)
result.duration_seconds = time.time() - start_time
logger.error(f"✗ Deployment to {task.device_id} failed: {e}")
finally:
# Disconnect
self.driver.disconnect(task.device_id)
# Store in history
self.deployment_history.append(result)
return result
def deploy_multiple_devices(self, tasks: List[DeploymentTask],
parallel: bool = False,
max_failures: Optional[int] = None) -> List[DeploymentResult]:
"""
Deploy configuration to multiple devices.
Args:
tasks: List of deployment tasks
parallel: If True, use parallel execution (Ray if available)
max_failures: Stop if this many devices fail (None = deploy all)
Returns:
List of DeploymentResults
"""
logger.info(f"Deploying to {len(tasks)} devices (parallel={parallel})")
if parallel and self.use_ray and self.ray_executor:
return self._deploy_parallel_ray(tasks, max_failures)
elif parallel:
logger.warning("Parallel mode requested but Ray not available - using serial")
# Serial deployment
results = []
failures = 0
for task in tasks:
result = self.deploy_single_device(task)
results.append(result)
if result.status == DeploymentStatus.FAILED:
failures += 1
if max_failures and failures >= max_failures:
logger.error(f"Max failures ({max_failures}) reached - stopping deployment")
# Mark remaining as pending
remaining = len(tasks) - len(results)
logger.warning(f"Skipping {remaining} remaining devices")
break
return results
def _deploy_parallel_ray(self, tasks: List[DeploymentTask],
max_failures: Optional[int]) -> List[DeploymentResult]:
"""Deploy using Ray parallel execution"""
logger.info(f"Using Ray to deploy to {len(tasks)} devices in parallel")
# This would integrate with Ray executor
# For now, fall back to serial
logger.warning("Ray parallel deployment not yet implemented - using serial")
return self.deploy_multiple_devices(tasks, parallel=False, max_failures=max_failures)
def _run_validation_check(self, device_id: str, check: str) -> bool:
"""
Run validation check on device.
Args:
device_id: Device identifier
check: Check command or test
Returns:
True if check passes
"""
try:
# Parse check type
if check.startswith("ping:"):
# Ping test
target = check.split(":", 1)[1]
result = self.driver.send_command(device_id, f"ping {target}")
return result.success and "success" in result.output.lower()
elif check.startswith("interface:"):
# Interface status check
interface = check.split(":", 1)[1]
result = self.driver.send_command(device_id, f"show interface {interface}")
return result.success and "up" in result.output.lower()
elif check.startswith("command:"):
# Generic command check (success = command runs without error)
command = check.split(":", 1)[1]
result = self.driver.send_command(device_id, command)
return result.success
else:
# Default: run as show command
result = self.driver.send_command(device_id, check)
return result.success
except Exception as e:
logger.error(f"Validation check '{check}' failed on {device_id}: {e}")
return False
def generate_and_deploy(self, device: Any, network_context: Dict[str, Any],
credentials: Dict[str, str],
dry_run: bool = False,
pre_checks: Optional[List[str]] = None,
post_checks: Optional[List[str]] = None) -> DeploymentResult:
"""
Generate configuration from template and deploy to device.
Args:
device: Device object from NetworkModel
network_context: Network-level context (VLANs, routing, etc.)
credentials: Device credentials (username, password)
dry_run: If True, generate and validate but don't deploy
pre_checks: Optional pre-deployment validation checks
post_checks: Optional post-deployment validation checks
Returns:
DeploymentResult
"""
device_name = device.get('name') if isinstance(device, dict) else device.name
logger.info(f"Generating and deploying config for {device_name}")
# 1. Generate configuration from template
config = self.template_engine.generate_device_config(device, network_context)
# 2. Map device vendor to driver type (handle both dict and object)
vendor = device.get('vendor') if isinstance(device, dict) else device.vendor
model = device.get('model', '') if isinstance(device, dict) else getattr(device, 'model', '')
device_type = self._map_vendor_to_device_type(vendor, model)
# 3. Create deployment task
device_name = device.get('name') if isinstance(device, dict) else device.name
mgmt_ip = device.get('mgmt_ip') if isinstance(device, dict) else device.mgmt_ip
task = DeploymentTask(
device_id=device_name,
device_type=device_type,
hostname=mgmt_ip,
username=credentials.get('username', 'admin'),
password=credentials.get('password', 'admin'),
config=config,
dry_run=dry_run,
pre_checks=pre_checks or [],
post_checks=post_checks or []
)
# 4. Deploy
return self.deploy_single_device(task)
def _map_vendor_to_device_type(self, vendor: str, model: str) -> DeviceType:
"""Map vendor/model to DeviceType"""
vendor_lower = vendor.lower()
model_lower = model.lower()
if 'cisco' in vendor_lower:
if 'nexus' in model_lower or 'nxos' in model_lower:
return DeviceType.CISCO_NXOS
elif 'xe' in model_lower or 'asr' in model_lower or 'isr' in model_lower:
return DeviceType.CISCO_XE
else:
return DeviceType.CISCO_IOS
elif 'arista' in vendor_lower:
return DeviceType.ARISTA_EOS
elif 'juniper' in vendor_lower:
return DeviceType.JUNIPER_JUNOS
else:
logger.warning(f"Unknown vendor {vendor}, defaulting to Cisco IOS")
return DeviceType.CISCO_IOS
def get_deployment_summary(self) -> Dict[str, Any]:
"""Get summary of deployment history"""
total = len(self.deployment_history)
if total == 0:
return {
'total_deployments': 0,
'success_count': 0,
'failed_count': 0,
'rolled_back_count': 0,
'success_rate': 0.0
}
success = sum(1 for r in self.deployment_history if r.status == DeploymentStatus.SUCCESS)
failed = sum(1 for r in self.deployment_history if r.status == DeploymentStatus.FAILED)
rolled_back = sum(1 for r in self.deployment_history if r.rolled_back)
return {
'total_deployments': total,
'success_count': success,
'failed_count': failed,
'rolled_back_count': rolled_back,
'success_rate': (success / total) * 100 if total > 0 else 0.0,
'avg_duration': sum(r.duration_seconds for r in self.deployment_history) / total,
'latest_deployments': [
{
'device_id': r.device_id,
'status': r.status.value,
'timestamp': r.timestamp.isoformat(),
'duration': r.duration_seconds
}
for r in self.deployment_history[-10:] # Last 10
]
}
def verify_deployment(self, device_id: str, expected_config_snippet: Optional[str] = None) -> Dict[str, Any]:
"""
Verify deployment was successful.
Args:
device_id: Device identifier
expected_config_snippet: Optional config snippet to verify is present
Returns:
Verification results
"""
results = {
'device_id': device_id,
'connectivity': False,
'config_retrieved': False,
'snippet_found': None,
'facts': None,
'errors': []
}
try:
# Check if we have a connection
if device_id not in self.driver.connections:
results['errors'].append("Device not connected")
return results
# Test connectivity
results['connectivity'] = self.driver.verify_connectivity(device_id)
if not results['connectivity']:
results['errors'].append("Device not responsive")
return results
# Get current config
config = self.driver.get_config(device_id, "running")
results['config_retrieved'] = config is not None
if not results['config_retrieved']:
results['errors'].append("Could not retrieve configuration")
return results
# Check for expected snippet
if expected_config_snippet:
results['snippet_found'] = expected_config_snippet in config
if not results['snippet_found']:
results['errors'].append(f"Expected config snippet not found: {expected_config_snippet[:50]}...")
# Get device facts
facts = self.driver.get_device_facts(device_id)
results['facts'] = facts
except Exception as e:
results['errors'].append(str(e))
logger.error(f"Verification failed for {device_id}: {e}")
return results
def cleanup(self):
"""Disconnect all devices and cleanup resources"""
self.driver.disconnect_all()
if self.ray_executor:
self.ray_executor.shutdown()
logger.info("DeploymentEngine cleanup complete")