maqet 0.0.1.3__py3-none-any.whl → 0.0.5__py3-none-any.whl

maqet/managers/vm_manager.py

@@ -0,0 +1,683 @@
+"""
+VM Manager
+
+Manages VM lifecycle operations: add, start, stop, remove, list.
+Extracted from Maqet class to follow Single Responsibility Principle.
+"""
+
+import os
+import time
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Union
+
+from ..config import ConfigMerger
+from ..constants import Intervals, Timeouts
+from ..exceptions import (
+    ConfigurationError,
+    RunnerSpawnError,
+    StateError,
+    VMAlreadyExistsError,
+    VMLifecycleError,
+    VMNotFoundError,
+    VMNotRunningError,
+    VMStartError,
+    VMStopError,
+)
+from ..logger import LOG
+from ..state import StateManager, VMInstance
+
+# Legacy exception alias (backward compatibility)
+VMManagerError = VMLifecycleError
+ConfigError = ConfigurationError
+StateManagerError = StateError
+
+
+class VMManager:
+    """
+    Manages VM lifecycle operations.
+
+    Responsibilities:
+    - Create VMs (add)
+    - Start VMs (spawn runner processes)
+    - Stop VMs (via IPC or process kill)
+    - Remove VMs (from database)
+    - List VMs
+    - Clean up dead processes
+    """
+
+    def __init__(self, state_manager: StateManager, config_parser):
+        """
+        Initialize VM manager.
+
+        Args:
+            state_manager: State management instance
+            config_parser: Configuration parser instance
+        """
+        self.state_manager = state_manager
+        self.config_parser = config_parser
+        LOG.debug("VMManager initialized")
+
+    def add(
+        self,
+        config: Optional[Union[str, List[str]]] = None,
+        name: Optional[str] = None,
+        empty: bool = False,
+        **kwargs,
+    ) -> str:
+        """
+        Create a new VM from configuration file(s) or parameters.
+
+        Args:
+            config: Path to YAML configuration file, or list of config
+                files for deep-merge
+            name: VM name (auto-generated if not provided)
+            empty: Create empty VM without any configuration (won't be
+                startable until configured)
+            **kwargs: Additional VM configuration parameters
+
+        Returns:
+            VM instance ID
+
+        Raises:
+            VMManagerError: If VM creation fails
+
+        Examples:
+            Single config: add(config="vm.yaml", name="myvm")
+            Multiple configs: add(
+                config=["base.yaml", "custom.yaml"], name="myvm"
+            )
+            Config + params: add(config="base.yaml", memory="8G", cpu=4)
+            Empty VM: add(name="placeholder-vm", empty=True)
+        """
+        try:
+            # Layer 2: Extract client working directory for path resolution
+            client_cwd = kwargs.pop("_client_cwd", None)
+
+            # Handle empty VM creation
+            if empty:
+                # Validate that no config or kwargs are provided with --empty
+                if config:
+                    raise VMManagerError(
+                        "Cannot specify config files with --empty flag"
+                    )
+                if kwargs:
+                    raise VMManagerError(
+                        "Cannot specify configuration parameters "
+                        "with --empty flag"
+                    )
+
+                # Create completely empty config
+                config_data = {}
+                config_file = None
+
+                # Generate name if not provided
+                if not name:
+                    # Generate unique name using UUID without creating temp VM
+                    import uuid
+                    unique_suffix = str(uuid.uuid4()).split('-')[-1][:8]
+                    name = f"empty-vm-{unique_suffix}"
+
+                # Skip validation for empty VMs
+                # Create VM in database with empty config
+                vm_id = self.state_manager.create_vm(
+                    name, config_data, config_file
+                )
+
+                return vm_id
+
+            # Normal VM creation path
+            # Load and deep-merge configuration files
+            if config:
+                config_data = ConfigMerger.load_and_merge_files(config)
+                if isinstance(config, str):
+                    config_file = config
+                elif config:
+                    config_file = config[0]
+                else:
+                    config_file = None
+            else:
+                config_data = {}
+                config_file = None
+
+            # Merge kwargs with config data (kwargs take precedence)
+            if kwargs:
+                config_data = ConfigMerger.deep_merge(config_data, kwargs)
+
+            # Handle name priority: CLI --name > config name > auto-generated
+            if not name:
+                # Check if name is present in merged config
+                name = config_data.get("name")
+
+            # Always remove name from config_data as it's VM metadata, not QEMU
+            # config
+            if "name" in config_data:
+                config_data = {
+                    k: v for k, v in config_data.items() if k != "name"
+                }
+
+            # Generate name if still not provided
+            if not name:
+                # Generate unique name using UUID without creating temp VM
+                import uuid
+                unique_suffix = str(uuid.uuid4()).split('-')[-1][:8]
+                name = f"vm-{unique_suffix}"
+
+            # Validate the final merged configuration
+            config_data = self.config_parser.validate_config(config_data)
+
+            # Create VM in database
+            vm_id = self.state_manager.create_vm(
+                name, config_data, config_file
+            )
+
+            return vm_id
+
+        # Specific exception handlers for better error messages
+        except FileNotFoundError as e:
+            raise VMManagerError(
+                f"Configuration file not found: {e.filename}. "
+                f"Check that the file path is correct."
+            )
+        except PermissionError as e:
+            raise VMManagerError(
+                f"Permission denied accessing configuration file: {
+                    e.filename}. "
+                f"Check file permissions and ownership."
+            )
+        except ConfigError as e:
+            raise VMManagerError(f"Configuration error: {e}")
+        except StateManagerError as e:
+            raise VMManagerError(f"Database error: {e}")
+        except Exception as e:
+            # Last resort - log unexpected errors with context
+            LOG.error(
+                f"Unexpected error creating VM: {type(e).__name__}: {e}",
+                exc_info=True,
+            )
+            raise VMManagerError(f"Failed to create VM: {e}")
+
+    def start(self, vm_id: str) -> VMInstance:
+        """
+        Start a virtual machine by spawning a detached VM runner process.
+
+        Changes from previous implementation:
+        - No longer manages Machine directly
+        - Spawns VM runner process that manages QEMU lifecycle
+        - VM runner survives CLI exit
+        - Returns immediately after runner is ready
+
+        Args:
+            vm_id: VM identifier (name or ID)
+
+        Returns:
+            VM instance information
+
+        Raises:
+            VMManagerError: If VM start fails
+        """
+        try:
+            # Get VM from database
+            vm = self.state_manager.get_vm(vm_id)
+            if not vm:
+                raise VMManagerError(f"VM '{vm_id}' not found")
+
+            # Check if VM is already running
+            if vm.status == "running":
+                # Check if runner process is actually alive
+                from ..process_spawner import is_runner_alive
+
+                if vm.runner_pid and is_runner_alive(vm.runner_pid):
+                    raise VMManagerError(
+                        f"VM '{vm_id}' is already running "
+                        f"(runner PID: {vm.runner_pid})"
+                    )
+                else:
+                    # Stale state - clean up and continue
+                    LOG.warning(
+                        f"VM '{vm_id}' has stale 'running' status, cleaning up"
+                    )
+                    self.state_manager.update_vm_status(
+                        vm_id, "stopped", runner_pid=None, socket_path=None
+                    )
+
+            # Check if VM has required configuration
+            if not vm.config_data or not vm.config_data.get("binary"):
+                raise VMManagerError(
+                    f"VM '{vm_id}' cannot be started: missing required "
+                    f"configuration. Use 'maqet apply {vm_id} "
+                    f"--config <config.yaml>' to add configuration."
+                )
+
+            # Spawn VM runner process
+            from ..process_spawner import (
+                spawn_vm_runner,
+                wait_for_vm_ready,
+                get_socket_path,
+            )
+
+            try:
+                # Get database path for runner
+                db_path = self.state_manager.xdg.database_path
+
+                runner_pid = spawn_vm_runner(vm.id, db_path, timeout=Timeouts.PROCESS_SPAWN)
+                LOG.info(f"Spawned VM runner process for '{vm_id}' (PID: {runner_pid})")
+            except Exception as e:
+                raise VMManagerError(f"Failed to spawn VM runner: {e}")
+
+            # Wait for VM runner to be ready (socket available)
+            socket_path = get_socket_path(vm.id)
+            ready = wait_for_vm_ready(vm.id, socket_path, timeout=Timeouts.VM_START)
+
+            if not ready:
+                # Runner process started but socket not available - cleanup
+                from ..process_spawner import kill_runner
+
+                kill_runner(runner_pid, force=True)
+                raise VMManagerError("VM runner did not become ready within timeout")
+
+            # Verify VM is actually running (runner updated DB)
+            vm_updated = self.state_manager.get_vm(vm_id)
+            if vm_updated.status != "running":
+                raise VMManagerError(
+                    f"VM runner started but VM status is '{vm_updated.status}'"
+                )
+
+            return vm_updated
+
+        except Exception as e:
+            raise VMManagerError(f"Failed to start VM '{vm_id}': {e}")
+
+    def stop(
+        self, vm_id: str, force: bool = False, timeout: int = 30
+    ) -> VMInstance:
+        """
+        Stop a VM by sending stop command to VM runner or killing runner process.
+
+        Changes from previous implementation:
+        - No longer uses Machine.stop() directly
+        - Sends stop command via IPC (graceful)
+        - Falls back to killing runner process if IPC fails
+
+        Args:
+            vm_id: VM identifier (name or ID)
+            force: If True, kill runner immediately (SIGKILL).
+                   If False, graceful shutdown (SIGTERM)
+            timeout: Timeout for graceful shutdown (ignored for now)
+
+        Returns:
+            VM instance information
+
+        Raises:
+            VMManagerError: If VM stop fails
+        """
+        try:
+            # Get VM from database
+            vm = self.state_manager.get_vm(vm_id)
+            if not vm:
+                raise VMManagerError(f"VM '{vm_id}' not found")
+
+            # Check if VM is running
+            if vm.status != "running":
+                LOG.info(f"VM '{vm_id}' is not running (status: {vm.status})")
+                # Ensure status is "stopped" (not "created", etc.)
+                if vm.status != "stopped":
+                    self.state_manager.update_vm_status(
+                        vm_id, "stopped", pid=None, runner_pid=None, socket_path=None
+                    )
+                    vm = self.state_manager.get_vm(vm_id)
+                return vm
+
+            from ..process_spawner import is_runner_alive, kill_runner
+            from ..ipc.runner_client import RunnerClient, RunnerClientError
+
+            # Verify runner process exists
+            if not vm.runner_pid or not is_runner_alive(vm.runner_pid):
+                # Runner missing/dead - check for orphaned QEMU process
+                LOG.warning(
+                    f"VM '{vm_id}' runner process not found, checking for orphaned QEMU"
+                )
+
+                # Fallback: Kill orphaned QEMU process if it exists
+                if vm.pid:
+                    try:
+                        # Check if QEMU process is alive
+                        os.kill(vm.pid, 0)  # Signal 0 checks existence
+
+                        # QEMU is alive - kill it
+                        LOG.warning(
+                            f"Found orphaned QEMU process (PID {vm.pid}), "
+                            f"terminating it"
+                        )
+                        if force:
+                            os.kill(vm.pid, 9)  # SIGKILL
+                        else:
+                            os.kill(vm.pid, 15)  # SIGTERM
+
+                        # Wait briefly for process to die
+                        time.sleep(0.5)
+
+                    except ProcessLookupError:
+                        # QEMU process already dead, no action needed
+                        LOG.debug(f"QEMU process {vm.pid} already dead")
+                    except PermissionError:
+                        LOG.error(
+                            f"Permission denied when killing QEMU process {vm.pid}"
+                        )
+                    except Exception as e:
+                        LOG.error(f"Failed to kill QEMU process {vm.pid}: {e}")
+
+                # Clean up DB
+                self.state_manager.update_vm_status(
+                    vm_id, "stopped", pid=None, runner_pid=None, socket_path=None
+                )
+                return self.state_manager.get_vm(vm_id)
+
+            # Try graceful stop via IPC (if not force)
+            if not force:
+                client = RunnerClient(vm.id, self.state_manager)
+
+                try:
+                    result = client.send_command("stop", timeout=timeout)
+                    LOG.info(f"VM '{vm_id}' stopped gracefully via IPC")
+
+                    # Wait briefly for runner to exit and update DB
+                    time.sleep(Intervals.CLEANUP_WAIT)
+
+                    # Verify status updated
+                    vm_updated = self.state_manager.get_vm(vm_id)
+                    return vm_updated
+
+                except RunnerClientError as e:
+                    LOG.warning(
+                        f"IPC stop failed for '{vm_id}': {e}, "
+                        f"falling back to SIGTERM"
+                    )
+
+            # Fallback: Kill runner process directly
+            LOG.info(
+                f"Killing VM runner process for '{vm_id}' "
+                f"(PID: {vm.runner_pid}, force={force})"
+            )
+            killed = kill_runner(vm.runner_pid, force=force)
+
+            if killed:
+                # Wait briefly for cleanup
+                time.sleep(Intervals.CLEANUP_WAIT)
+
+                # Verify DB updated (runner should clean up on exit)
+                vm_updated = self.state_manager.get_vm(vm_id)
+                if vm_updated.status == "running":
+                    # Runner didn't clean up - do it manually
+                    self.state_manager.update_vm_status(
+                        vm_id, "stopped", runner_pid=None, socket_path=None
+                    )
+                    vm_updated = self.state_manager.get_vm(vm_id)
+
+                return vm_updated
+            else:
+                raise VMManagerError(f"Failed to kill runner process {vm.runner_pid}")
+
+        except Exception as e:
+            raise VMManagerError(f"Failed to stop VM '{vm_id}': {e}")
+
+    def remove(
+        self,
+        vm_id: Optional[str] = None,
+        force: bool = False,
+        all: bool = False,
+        clean_storage: bool = False,
+    ) -> bool:
+        """
+        Remove a virtual machine completely.
+
+        Args:
+            vm_id: VM identifier (name or ID)
+            force: Force removal even if VM is running
+            all: Remove all virtual machines
+            clean_storage: Also delete associated storage files
+
+        Returns:
+            True if removed successfully
+
+        Raises:
+            VMManagerError: If VM removal fails
+        """
+        try:
+            # Validate arguments
+            if all and vm_id:
+                raise VMManagerError("Cannot specify both vm_id and --all flag")
+            if not all and not vm_id:
+                raise VMManagerError("Must specify either vm_id or --all flag")
+
+            # Handle bulk removal
+            if all:
+                return self._remove_all_vms(force, clean_storage)
+
+            # Handle single VM removal
+            return self._remove_single_vm(vm_id, force, clean_storage)
+
+        except Exception as e:
+            raise VMManagerError(f"Failed to remove VM(s): {e}")
+
+    def _remove_single_vm(
+        self, vm_id: str, force: bool, clean_storage: bool = False
+    ) -> bool:
+        """Remove a single VM."""
+        # Get VM from database
+        vm = self.state_manager.get_vm(vm_id)
+        if not vm:
+            raise VMManagerError(f"VM '{vm_id}' not found")
+
+        # Stop VM if running
+        if vm.status == "running":
+            if not force:
+                raise VMManagerError(
+                    f"VM '{vm_id}' is running. Use --force to remove "
+                    f"running VMs"
+                )
+            self.stop(vm_id, force=True)
+
+        # Clean up storage files if requested
+        if clean_storage:
+            storage_configs = vm.config_data.get("storage", [])
+            for storage in storage_configs:
+                if "file" in storage:
+                    storage_path = Path(storage["file"])
+                    if storage_path.exists():
+                        try:
+                            LOG.info(f"Removing storage file: {storage_path}")
+                            storage_path.unlink()
+                        except OSError as e:
+                            LOG.warning(
+                                f"Failed to remove storage file {
+                                    storage_path}: {e}"
+                            )
+
+        # Remove from database
+        removed = self.state_manager.remove_vm(vm_id)
+        if not removed:
+            raise VMManagerError(f"Failed to remove VM '{vm_id}' from database")
+
+        return True
+
+    def _remove_all_vms(
+        self, force: bool, clean_storage: bool = False
+    ) -> bool:
+        """Remove all VMs with confirmation."""
+        # Get all VMs
+        all_vms = self.state_manager.list_vms()
+
+        if not all_vms:
+            print("No virtual machines found.")
+            return True
+
+        # Display VMs that will be removed
+        print(f"Found {len(all_vms)} virtual machine(s) to remove:")
+        print()
+
+        # Create table header
+        header = f"{'NAME':<20} {'STATUS':<10} {'PID':<8}"
+        separator = "-" * 40
+        print(header)
+        print(separator)
+
+        running_count = 0
+        for vm in all_vms:
+            pid_str = str(vm.pid) if vm.pid else "-"
+            print(f"{vm.name:<20} {vm.status:<10} {pid_str:<8}")
+            if vm.status == "running":
+                running_count += 1
+
+        print()
+
+        # Show warning for running VMs
+        if running_count > 0 and not force:
+            print(
+                f"WARNING: {running_count} VM(s) are currently running "
+                f"and will be forcefully stopped."
+            )
+            print("Use --force to skip this warning in the future.")
+            print()
+
+        # Confirmation prompt
+        try:
+            response = (
+                input(
+                    f"Are you sure you want to remove ALL {len(all_vms)} "
+                    f"virtual machines? [y/N]: "
+                )
+                .strip()
+                .lower()
+            )
+        except (EOFError, KeyboardInterrupt):
+            print("\nOperation cancelled.")
+            return False
+
+        if response not in ["y", "yes"]:
+            print("Operation cancelled.")
+            return False
+
+        # Remove all VMs
+        removed_count = 0
+        failed_count = 0
+
+        print()
+        print("Removing virtual machines...")
+
+        for vm in all_vms:
+            try:
+                # Stop VM if running
+                if vm.status == "running":
+                    try:
+                        self.stop(vm.id, force=True)
+                        print(f"  Stopped VM: {vm.name}")
+                    except Exception as e:
+                        print(f"  Warning: Failed to stop VM '{vm.name}': {e}")
+
+                # Remove from database
+                removed = self.state_manager.remove_vm(vm.id)
+                if removed:
+                    print(f"  Removed VM: {vm.name}")
+                    removed_count += 1
+                else:
+                    print(f"  Failed to remove VM: {vm.name}")
+                    failed_count += 1
+
+            except Exception as e:
+                print(f"  Error removing VM '{vm.name}': {e}")
+                failed_count += 1
+
+        print()
+        print(
+            f"Removal complete: {removed_count} removed, "
+            f"{failed_count} failed"
+        )
+
+        if failed_count > 0:
+            raise VMManagerError(f"Failed to remove {failed_count} VM(s)")
+
+        return True
+
+    def list_vms(self, status: Optional[str] = None) -> List[VMInstance]:
+        """
+        List virtual machines.
+
+        Args:
+            status: Filter by status ('running', 'stopped', 'created',
+                'failed')
+
+        Returns:
+            List of VM instances
+        """
+        vms = self.state_manager.list_vms(status_filter=status)
+
+        # Check process status and update if needed
+        for vm in vms:
+            if vm.status == "running" and vm.pid:
+                if not self.state_manager._is_process_alive(vm.pid):
+                    # Process doesn't exist, update status
+                    self.state_manager.update_vm_status(
+                        vm.id, "stopped", pid=None
+                    )
+                    vm.status = "stopped"
+                    vm.pid = None
+
+        return vms
+
+    def cleanup_dead_processes(self) -> List[str]:
+        """
+        Check for VMs with running status but dead runner processes.
+        Update DB to reflect reality.
+
+        This runs on VMManager initialization to clean up stale state from
+        crashed runners or improperly terminated VMs.
+
+        Returns:
+            List of VM IDs that were cleaned up
+        """
+        from ..process_spawner import is_runner_alive
+
+        cleaned = []
+
+        # Get all VMs marked as running
+        all_vms = self.state_manager.list_vms()
+        running_vms = [vm for vm in all_vms if vm.status == "running"]
+
+        for vm in running_vms:
+            if not vm.runner_pid or not is_runner_alive(vm.runner_pid):
+                LOG.warning(
+                    f"VM '{vm.name}' marked as running but runner process "
+                    f"(PID: {vm.runner_pid}) is dead"
+                )
+
+                # Check for orphaned QEMU process
+                if vm.pid:
+                    try:
+                        # Check if QEMU is alive
+                        os.kill(vm.pid, 0)
+
+                        # QEMU is alive - kill it
+                        LOG.warning(
+                            f"Found orphaned QEMU process (PID {vm.pid}) for VM "
+                            f"'{vm.name}', terminating it"
+                        )
+                        os.kill(vm.pid, 9)  # SIGKILL
+                        time.sleep(0.5)
+
+                    except ProcessLookupError:
+                        # QEMU already dead
+                        LOG.debug(f"QEMU process {vm.pid} already dead")
+                    except PermissionError:
+                        LOG.error(
+                            f"Permission denied when killing QEMU process {vm.pid}"
+                        )
+                    except Exception as e:
+                        LOG.error(f"Failed to kill QEMU process {vm.pid}: {e}")
+
+                # Update DB
+                self.state_manager.update_vm_status(
+                    vm.name, "stopped", pid=None, runner_pid=None, socket_path=None
+                )
+                cleaned.append(vm.id)
+
+        return cleaned