matrice-compute 0.1.35__py3-none-any.whl → 0.1.36__py3-none-any.whl

matrice_compute/action_instance.py

@@ -10,7 +10,6 @@ import signal
 import urllib.request
 from matrice_compute.instance_utils import (
     get_gpu_with_sufficient_memory_for_action,
-    get_gpu_config_for_deployment,
     get_decrypted_access_key_pair,
     get_max_file_system,
     get_best_service_ip_and_network,
@@ -27,10 +26,6 @@ from matrice_common.utils import log_errors
 class ActionInstance:
     """Base class for tasks that run in Action containers."""
 
-    # Class-level dictionary to track deployed services and their ports
-    # Key: _idService, Value: {"triton_ports": "port1,port2,port3", "is_first": False}
-    _deployed_services = {}
-
     def __init__(self, scaling: Scaling, action_info: dict):
         """Initialize an action instance.
 
@@ -90,67 +85,6 @@ class ActionInstance:
             raise ValueError(f"Unknown action type: {self.action_type}")
         self.task = self.actions_map[self.action_type]
 
-    @classmethod
-    def is_first_deployment_for_service(cls, service_id):
-        """Check if this is the first deployment for a given service.
-
-        Args:
-            service_id (str): Service ID (_idService)
-
-        Returns:
-            bool: True if this is the first deployment, False otherwise
-        """
-        if not service_id:
-            return False
-        return service_id not in cls._deployed_services
-
-    @classmethod
-    def get_or_create_triton_ports(cls, service_id, scaling_instance):
-        """Get existing TRITON_PORTS for a service or create new ones.
-
-        Args:
-            service_id (str): Service ID (_idService)
-            scaling_instance: Scaling instance to get open ports
-
-        Returns:
-            str: Comma-separated string of 3 port numbers (e.g., "8001,8002,8003")
-        """
-        if not service_id:
-            # No service_id, generate new ports
-            port1 = scaling_instance.get_open_port()
-            port2 = scaling_instance.get_open_port()
-            port3 = scaling_instance.get_open_port()
-            return f"{port1},{port2},{port3}"
-
-        # Check if ports already exist for this service
-        if service_id in cls._deployed_services:
-            triton_ports = cls._deployed_services[service_id]["triton_ports"]
-            logging.info(
-                "Reusing TRITON_PORTS for service %s: %s",
-                service_id,
-                triton_ports
-            )
-            return triton_ports
-
-        # First deployment: generate new ports and store them
-        port1 = scaling_instance.get_open_port()
-        port2 = scaling_instance.get_open_port()
-        port3 = scaling_instance.get_open_port()
-        triton_ports = f"{port1},{port2},{port3}"
-
-        # Store for future use
-        cls._deployed_services[service_id] = {
-            "triton_ports": triton_ports,
-            "is_first": False
-        }
-
-        logging.info(
-            "First deployment for service %s - generated TRITON_PORTS: %s",
-            service_id,
-            triton_ports
-        )
-        return triton_ports
-
     @log_errors(default_return={}, raise_exception=True, log_error=False)
     def _init_credentials(self):
         """Initialize Matrice credentials.
@@ -297,7 +231,7 @@ class ActionInstance:
                         getattr(self, "action_record_id", "unknown"),
                     )
                 else:
-                    logging.info(
+                    logging.debug(
                         "No additional logs to send for action %s",
                         getattr(self, "action_record_id", "unknown"),
                     )
@@ -352,13 +286,13 @@ class ActionInstance:
         ).get("gpuMemory", 0)
 
         logging.info(
-            "Action %s requires GPU with %d MB memory - selecting GPU(s) with most free memory",
+            "Action %s requires GPU with %d MB memory - selecting best-fit GPU(s)",
             action_id,
             required_memory
         )
 
         try:
-            # Get the GPU(s) with most free memory that have sufficient memory
+            # Get the best-fit GPU(s) with sufficient memory
             gpu_indices = get_gpu_with_sufficient_memory_for_action(
                 action_details=action_details
             )
@@ -412,7 +346,6 @@ class ActionInstance:
         destination_workspace_path: str = "/usr/src/workspace",
         docker_workdir: str = "",
         extra_pkgs: list = [],
-        container_name: str = "",
     ):
         """Build base Docker command with common options.
 
@@ -427,7 +360,6 @@ class ActionInstance:
             destination_workspace_path (str): Container workspace path
             docker_workdir (str): Docker working directory
             extra_pkgs (list): List of extra packages to install
-            container_name (str): Docker container name (format: {action_type}_{action_id})
         Returns:
             str: Base Docker command
         """
@@ -492,20 +424,17 @@ class ActionInstance:
             ]
         )
 
-        # Build container name option if provided
-        name_option = f"--name {container_name}" if container_name else ""
-
         # if the service provider is local, then put --restart unless-stopped
         if os.environ.get("SERVICE_PROVIDER") in ("local", "LOCAL"):
             env_exports += " && export DOCKER_RESTART_POLICY='--restart unless-stopped' "
 
         cmd_parts = [
-            f"docker run -d {use_gpu} ",
-            name_option,
+            f"docker run {use_gpu} ",
             network_config,
             *[f"-e {key}={shlex.quote(str(value))}" for key, value in env_vars.items()],
             *volumes,
             # Container configuration and startup commands
+            f"--cidfile ./{self.action_record_id}.cid ",
             f"--shm-size=30G --pull=always {shlex.quote(self.docker_container)}",
             f'/bin/bash -c "cd {docker_workdir} && '
             f"{env_exports} && "
@@ -893,34 +822,6 @@ class ActionInstance:
                 job_params=action_details["jobParams"],
             )
 
-    @staticmethod
-    def container_exists(container_id: str) -> bool:
-        """Check if a Docker container exists.
-
-        Args:
-            container_id (str): Container ID or name to check
-
-        Returns:
-            bool: True if container exists, False otherwise
-        """
-        if not container_id:
-            return False
-        try:
-            result = subprocess.run(
-                ["docker", "inspect", container_id],
-                capture_output=True,
-                text=True,
-                timeout=10
-            )
-            return result.returncode == 0
-        except Exception as e:
-            logging.warning(
-                "Error checking if container %s exists: %s",
-                container_id,
-                str(e)
-            )
-            return False
-
     @log_errors(raise_exception=True)
     def start_process(self, cmd, log_name):
         """Start the process and initialize logging.
@@ -935,54 +836,60 @@ class ActionInstance:
         self.cmd = cmd
         self.log_path = f"{self.get_log_path()}/{log_name}_{self.action_record_id}.txt"
 
-        # Run docker with -d flag to get container ID from stdout
-        process = subprocess.Popen(
-            shlex.split(self.cmd),
-            stdout=subprocess.PIPE,
-            stderr=subprocess.PIPE,
-            text=True,
-            env={**os.environ},
-        )
-        
-        # Use a longer timeout for docker run since --pull=always may need to
-        # download large images on first run. Default: 30 minutes (1800 seconds)
-        # Can be configured via DOCKER_START_TIMEOUT_SECONDS environment variable
-        docker_start_timeout = int(os.environ.get("DOCKER_START_TIMEOUT_SECONDS", 1800))
-        logging.info(
-            "Waiting for docker container to start for action %s (timeout: %d seconds)",
-            self.action_record_id,
-            docker_start_timeout,
-        )
-        stdout, stderr = process.communicate(timeout=docker_start_timeout)
+        with open(self.log_path, "wb") as out:
+            self.process = subprocess.Popen(
+                shlex.split(self.cmd),
+                stdout=out,
+                stderr=out,
+                env={**os.environ},
+                start_new_session=True,
+            )
         
-        if process.returncode != 0:
+        self.container_id = None
+
+        cid_file_path = f"./{self.action_record_id}.cid"
+        max_retries = 5
+        retry_delay = 1  # seconds
+        for attempt in range(max_retries):
+            try:
+                with open(cid_file_path, "r") as cid_file:
+                    container_id = cid_file.read().strip()
+                    self.container_id = container_id
+                    logging.info(
+                        "Started process for action %s with container ID: %s",
+                        self.action_record_id,
+                        self.container_id,
+                    )
+                    break
+            except FileNotFoundError:
+                logging.warning(
+                    "CID file not found for action %s, attempt %d/%d",
+                    self.action_record_id,
+                    attempt + 1,
+                    max_retries,
+                )
+                time.sleep(retry_delay)
+            except Exception as e:
+                logging.error(
+                    "Error reading CID file for action %s: %s",
+                    self.action_record_id,
+                    str(e),
+                )
+                time.sleep(retry_delay)
+        else:
             logging.error(
-                "Docker run failed for action %s: %s",
+                "Failed to read CID file for action %s after %d attempts",
                 self.action_record_id,
-                stderr,
+                max_retries,
             )
-            raise RuntimeError(f"Docker run failed: {stderr}")
-        
-        self.container_id = stdout.strip()
-        logging.info(
-            "Started container for action %s with ID: %s",
-            self.action_record_id,
-            self.container_id,
-        )
-        
-        # Start following container logs in background
-        self.process = subprocess.Popen(
-            ["docker", "logs", "-f", self.container_id],
-            stdout=open(self.log_path, "wb"),
-            stderr=subprocess.STDOUT,
-            start_new_session=True,
-        )
+            raise Exception("Failed to start process: CID file not found")
         
-        # Report container id to scaling service
+        # report container id to scaling service
         self.scaling.update_action_container_id(
             action_record_id=self.action_record_id,
             container_id=self.container_id,
         )
+    
 
     @log_errors(raise_exception=False)
     def start_logger(self):
@@ -1143,8 +1050,7 @@ def data_preparation_execute(
             "Started pulling Docker image with PID: %s",
             process.pid,
         )
-    container_name = f"data_prep_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, destination_workspace_path="/usr/src/app/workspace", docker_workdir="/usr/src/app/workspace", extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/data_preparation.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs, destination_workspace_path="/usr/src/app/workspace", docker_workdir="/usr/src/app/workspace", extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/data_preparation.py {self.action_record_id} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "data_preparation_log")
 
@@ -1173,8 +1079,7 @@ def data_processing_execute(self: ActionInstance):
         service="bg-job-scheduler",
         job_params=action["jobParams"],
     )
-    container_name = f"data_processing_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/main.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/main.py {self.action_record_id} "'
     logging.info("cmd: %s", cmd)
     self.start(cmd, "data_processing_log")
 
@@ -1187,8 +1092,7 @@ def data_split_execute(self: ActionInstance):
     if not action_details:
         return
     self.setup_action_requirements(action_details, work_fs, model_family="")
-    container_name = f"data_split_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/data_split.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/data_split.py {self.action_record_id} "'
     logging.info("cmd: %s", cmd)
     self.start(cmd, "data_split")
 
@@ -1203,8 +1107,7 @@ def dataset_annotation_execute(
     if not action_details:
         return
     self.setup_action_requirements(action_details, work_fs)
-    container_name = f"dataset_annotation_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/dataset_annotation.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/dataset_annotation.py {self.action_record_id} "'
     logging.info("cmd: %s", cmd)
     self.start(cmd, "dataset_annotation")
 
@@ -1219,8 +1122,7 @@ def dataset_augmentation_execute(
     if not action_details:
         return
     self.setup_action_requirements(action_details, work_fs)
-    container_name = f"dataset_augmentation_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/data_augmentation.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/data_augmentation.py {self.action_record_id} "'
     logging.info("cmd: %s", cmd)
     self.start(cmd, "dataset_augmentation")
 
@@ -1236,8 +1138,7 @@ def augmentation_server_creation_execute(
     if not action_details:
         return
     self.setup_action_requirements(action_details, work_fs)
-    container_name = f"augmentation_setup_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/aug_server.py {self.action_record_id} {external_port} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs, extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/aug_server.py {self.action_record_id} {external_port} "'
     logging.info("cmd: %s", cmd)
     self.start(cmd, "augmentation_setup")
 
@@ -1258,45 +1159,30 @@ def database_setup_execute(self: ActionInstance):
 
     project_id = action_details["_idProject"]
 
-    # Define container names with action_record_id for uniqueness
-    mongodb_container_name = f"database_setup_{self.action_record_id}"
-    qdrant_container_name = f"qdrant_{self.action_record_id}"
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for inference tracker: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "qdrant_setup")
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if both containers actually exist before trying to restart
-        mongodb_container_exists = ActionInstance.container_exists(existing_container_id)
-        qdrant_container_exists = ActionInstance.container_exists(qdrant_container_name)
+        #qdrant restart
+        qdrant_cmd = "docker restart qdrant"
+        self.start(qdrant_cmd, 'qdrant_setup')
 
-        if mongodb_container_exists and qdrant_container_exists:
-            logging.info(
-                "Using existing container ID for database setup: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "database_setup")
+        return
+    
 
-            # qdrant restart
-            qdrant_cmd = f"docker restart {qdrant_container_name}"
-            self.start(qdrant_cmd, "qdrant_setup")
-            return
-        else:
-            logging.warning(
-                "Container(s) not found (mongodb=%s, qdrant=%s). Creating new containers.",
-                mongodb_container_exists,
-                qdrant_container_exists
-            )
-            # Fall through to create new containers
+    dbPath =action_details["jobParams"].get("dbPath","/host/data/path/mongodb_data")
 
-    dbPath = action_details["jobParams"].get("dbPath", "/host/data/path/mongodb_data")
 
     # MongoDB container with --net=host (Port: 27020:27017)
     cmd = (
         f"docker run --pull=always --net=host "
-        f"--name {mongodb_container_name} "
-        f"-v matrice_myvol:/matrice_data "
         f"-v {dbPath}:{dbPath} "
+        f"--name database_setup_{self.action_record_id} "
         f"-v /var/run/docker.sock:/var/run/docker.sock "
         f"--cidfile ./{self.action_record_id}.cid "
         f"-e ACTION_RECORD_ID={self.action_record_id} "
@@ -1308,23 +1194,6 @@ def database_setup_execute(self: ActionInstance):
     )
     logging.info("Starting DB container (Port: 27020:27017): %s", cmd)
 
-    # Qdrant container with --net=host (Port: 6334)
-    qdrant_cmd = (
-        f"docker run -d --pull=always --net=host "
-        f"--name {qdrant_container_name} "
-        f"-v matrice_myvol:/matrice_data "
-        f"qdrant/qdrant:latest "
-    )
-    logging.info("Starting Qdrant container (Port: 6334): %s", qdrant_cmd)
-
-    # Start Qdrant container
-    qdrant_process = subprocess.Popen(
-        qdrant_cmd,
-        shell=True,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.PIPE,
-    )
-    logging.info("Qdrant container started successfully")
 
     # Docker Command run
     self.start(cmd, "database_setup") 
@@ -1344,32 +1213,23 @@ def facial_recognition_setup_execute(self: ActionInstance):
 
     self.setup_action_requirements(action_details)
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for facial recognition worker: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "facial_recognition_setup")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for facial recognition worker: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "facial_recognition_setup")
+        return
 
     # Facial recognition worker container with --net=host (Port: 8081)
-    container_name = f"facial_recognition_{self.action_record_id}"
     worker_cmd = (
         f"docker run -d --pull=always --net=host "
-        f"--name {container_name} "
-        f"--cidfile ./{self.action_record_id}.cid "
+        f"--name worker "
+         f"--cidfile ./{self.action_record_id}.cid "
         f"-v matrice_myvol:/matrice_data "
+        f"--cidfile ./{self.action_record_id}.cid "
         f'-e ENV="{os.environ.get("ENV", "prod")}" '
         f'-e MATRICE_SECRET_ACCESS_KEY="{self.matrice_secret_access_key}" '
         f'-e MATRICE_ACCESS_KEY_ID="{self.matrice_access_key_id}" '
@@ -1396,30 +1256,20 @@ def lpr_setup_execute(self: ActionInstance):
 
     self.setup_action_requirements(action_details)
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for LPR worker: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "lpr_setup")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for LPR worker: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "lpr_setup")
+        return
 
     # LPR worker container with --net=host (Port: 8082)
-    container_name = f"lpr_{self.action_record_id}"
     worker_cmd = (
         f"docker run -d --net=host --pull=always "
-        f"--name {container_name} "
+        f"--name lpr-worker "
         f"--cidfile ./{self.action_record_id}.cid "
         f"-v matrice_myvol:/matrice_data "
         f'-e ENV="{os.environ.get("ENV", "prod")}" '
@@ -1458,30 +1308,20 @@ def inference_ws_server_execute(self: ActionInstance):
     
     logging.info(f"Inference WebSocket server will use IP: {ws_host} on port 8102 (use_host_network={use_host_network})")
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for inference WebSocket server: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "inference_ws_server")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for inference WebSocket server: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "inference_ws_server")
+        return
 
     # Inference WebSocket server with --net=host (Port: 8102)
-    container_name = f"inference_ws_{self.action_record_id}"
     worker_cmd = (
         f"docker run -d --pull=always --net=host "
-        f"--name {container_name} "
+        f"--name inference "
         f"--cidfile ./{self.action_record_id}.cid "
         f'-e ENV="{os.environ.get("ENV", "prod")}" '
         f'-e MATRICE_SECRET_ACCESS_KEY="{self.matrice_secret_access_key}" '
@@ -1517,30 +1357,20 @@ def fe_fs_streaming_execute(self: ActionInstance):
     
     logging.info(f"Frontend streaming will connect to WebSocket at: {ws_url}")
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for frontend streaming: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "fe_fs_streaming")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
-
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for frontend streaming: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "fe_fs_streaming")
+        return
+    
     # Frontend streaming with --net=host (Port: 3000)
-    container_name = f"fe_streaming_{self.action_record_id}"
     worker_cmd = (
         f"docker run -d --pull=always --net=host "
-        f"--name {container_name} "
+        f"--name fe_streaming "
         f"--cidfile ./{self.action_record_id}.cid "
         f"-v matrice_myvol:/matrice_data "
         f'-e ENV="{os.environ.get("ENV", "prod")}" '
@@ -1573,30 +1403,20 @@ def fe_analytics_service_execute(self: ActionInstance):
 
     project_id = action_details["_idProject"]
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for frontend analytics service: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "fe_analytics_service")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
-
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for frontend analytics service: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "fe_analytics_service")
+        return
+    
     # Frontend analytics service with --net=host (Port: 3001)
-    container_name = f"fe_analytics_{self.action_record_id}"
     worker_cmd = (
         f"docker run -d --pull=always --net=host "
-        f"--name {container_name} "
+        f"--name fe-analytics "
         f"--cidfile ./{self.action_record_id}.cid "
         f'-e NEXT_PUBLIC_DEPLOYMENT_ENV="{os.environ.get("ENV", "prod")}" '
         f'-e MATRICE_SECRET_ACCESS_KEY="{self.matrice_secret_access_key}" '
@@ -1629,8 +1449,7 @@ def synthetic_dataset_generation_execute(self: ActionInstance):
     else:
         return
     use_gpu = self.get_gpu_config(action_details)
-    container_name = f"dataset_generation_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs=work_fs, use_gpu=use_gpu, extra_env_vars=extra_env_vars, extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/synthetic_dataset_generation.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs=work_fs, use_gpu=use_gpu, extra_env_vars=extra_env_vars, extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/synthetic_dataset_generation.py {self.action_record_id} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "dataset_generation")
 
@@ -1651,8 +1470,7 @@ def synthetic_data_setup_execute(self: ActionInstance):
     else:
         return
     use_gpu = self.get_gpu_config(action_details)
-    container_name = f"synthetic_data_setup_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs=work_fs, use_gpu=use_gpu, extra_env_vars=extra_env_vars, extra_pkgs=["matrice_dataset"], container_name=container_name)} python3 /usr/src/app/data_generation.py {self.action_record_id} {external_port} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs=work_fs, use_gpu=use_gpu, extra_env_vars=extra_env_vars, extra_pkgs=["matrice_dataset"])} python3 /usr/src/app/data_generation.py {self.action_record_id} {external_port} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "synthetic_data_setup")
 
@@ -1689,60 +1507,31 @@ def redis_setup_execute(self: ActionInstance):
 
     redis_image = action_details["actionDetails"].get("redis_image", "redis:latest")
 
-    # Define container names with action_record_id for uniqueness
-    redis_container_name = f"redis_{self.action_record_id}"
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if both containers actually exist before trying to restart
-        management_container_exists = ActionInstance.container_exists(existing_container_id)
-        redis_container_exists = ActionInstance.container_exists(redis_container_name)
-
-        if management_container_exists and redis_container_exists:
-            logging.info(
-                "Using existing container ID for redis management: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "redis_setup")
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for redis management: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "redis_setup")
 
-            # Redis container restart
-            redis_restart_cmd = f"docker restart {redis_container_name}"
-            self.start(redis_restart_cmd, "redis")
-            return
-        else:
-            logging.warning(
-                "Container(s) not found (management=%s, redis=%s). Creating new containers.",
-                management_container_exists,
-                redis_container_exists
-            )
-            # Fall through to create new containers
+        # Redis container restart
+        redis_restart_cmd = "docker restart redis_container"
+        self.start(redis_restart_cmd, "redis")
 
+        return
+    
     # Redis container with --net=host (Port: 6379)
     redis_cmd = (
         f"docker run -d --net=host "
-        f"--name {redis_container_name} "
+        f"--name redis_container "
         f"--restart unless-stopped "
         f"{redis_image} "
-        f"redis-server --bind 0.0.0.0 "
-        f"--appendonly no "
-        f'--save "" '
-        f"--maxmemory 30gb "
-        f"--maxmemory-policy allkeys-lru "
-        f"--io-threads 4 "
-        f"--io-threads-do-reads yes "
-        f"--stream-node-max-bytes 8192 "
-        f"--stream-node-max-entries 1000 "
-        f"--hz 100 "
-        f"--tcp-backlog 2048 "
-        f"--timeout 0 "
-        f"--lazyfree-lazy-eviction yes "
-        f"--lazyfree-lazy-expire yes "
-        f"--lazyfree-lazy-server-del yes "
-        f"--activedefrag yes "
-        f"--requirepass {redis_password}"
+        f"redis-server --bind 0.0.0.0 --appendonly yes --requirepass {redis_password}"
     )
+    
     logging.info("Starting Redis container on %s:6379: %s", redis_host, redis_cmd)
     
     # Start Redis container first
@@ -1792,8 +1581,7 @@ def deploy_aggregator_execute(
     if not action_details:
         return
     self.setup_action_requirements(action_details, work_fs)
-    container_name = f"deploy_aggregator_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, container_name=container_name)} python3 /usr/src/app/deploy_aggregator.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd(work_fs)} python3 /usr/src/app/deploy_aggregator.py {self.action_record_id} "'
     logging.info("cmd: %s", cmd)
     self.start(cmd, "deploy_aggregator")
 
@@ -1809,10 +1597,6 @@ def model_deploy_execute(self: ActionInstance):
         return
     action_id = action_details["_id"]
     model_family = action_details["actionDetails"]["modelFamily"]
-
-    # Get the service ID to track deployments
-    service_id = action_details.get("_idService")
-
     self.setup_action_requirements(
         action_details,
         work_fs,
@@ -1820,29 +1604,17 @@ def model_deploy_execute(self: ActionInstance):
         action_id=action_id,
     )
 
-    # Check if this is the first deployment for this service
-    is_first_deployment = ActionInstance.is_first_deployment_for_service(service_id)
-
-    # Get GPU configuration (uses utility function with fail-safe fallback)
-    use_gpu = get_gpu_config_for_deployment(action_details, is_first_deployment)
-
-    logging.info(
-        "Action %s: Model deployment GPU config: %s (first_deployment=%s)",
-        action_id,
-        use_gpu if use_gpu else "CPU-only",
-        is_first_deployment
-    )
-
-    # Get or create TRITON_PORTS (uses utility method)
-    triton_ports = ActionInstance.get_or_create_triton_ports(service_id, self.scaling)
+    # Get GPU configuration based on requirements and availability
+    # This uses the best-fit algorithm to select the most appropriate GPU(s)
+    use_gpu = self.get_gpu_config(action_details)
 
-    extra_env_vars = {
-        "INTERNAL_PORT": internal_port,
-        "TRITON_PORTS": triton_ports
-    }
+    # Override: If GPU is required, use all available GPUs
+    gpuRequired = action_details["actionDetails"].get("gpuRequired", False)
+    if gpuRequired:
+        use_gpu = "--runtime=nvidia --gpus all"
 
-    container_name = f"model_deploy_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, mount_docker_sock=True, action_id=action_id, extra_env_vars=extra_env_vars, extra_pkgs=["matrice_inference", "matrice_analytics"], container_name=container_name)} python3 deploy.py {self.action_record_id} {external_port}"'
+    extra_env_vars = {"INTERNAL_PORT": internal_port}
+    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, mount_docker_sock=True, action_id=action_id, extra_env_vars=extra_env_vars, extra_pkgs=["matrice_inference", "matrice_analytics"])} python3 deploy.py {self.action_record_id} {external_port}"'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "deploy_log")
 
@@ -1865,27 +1637,17 @@ def model_train_execute(self: ActionInstance):
         action_id=action_id,
     )
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for training: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "train_log")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
-
-    container_name = f"model_train_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, action_id=action_id, model_key=model_key, container_name=container_name)} python3 train.py {self.action_record_id} "'
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for training: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "train_log")
+        return
+    
+    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, action_id=action_id, model_key=model_key)} python3 train.py {self.action_record_id} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "train_log")
 
@@ -1906,27 +1668,17 @@ def model_eval_execute(self: ActionInstance):
         model_family=model_family,
         action_id=action_id,
     )
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for evaluation: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "eval_log")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
-
-    container_name = f"model_eval_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, action_id=action_id, container_name=container_name)} python3 eval.py {self.action_record_id} "'
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for training: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "eval_log")
+        return
+    
+    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, action_id=action_id)} python3 eval.py {self.action_record_id} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "eval_log")
 
@@ -1950,27 +1702,17 @@ def model_export_execute(self: ActionInstance):
         model_family=model_family,
         action_id=action_id,
     )
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for export: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "export_log")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
-
-    container_name = f"model_export_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, action_id=action_id, container_name=container_name)} python3 export.py {self.action_record_id} "'
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for training: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "export_log")
+        return
+    
+    cmd = f'{self.get_base_docker_cmd(work_fs, use_gpu, action_id=action_id)} python3 export.py {self.action_record_id} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "export_log")
 
@@ -1986,8 +1728,7 @@ def image_build_execute(self: ActionInstance):
     action_id = action_details["_id"]
     internal_api_key = self.get_internal_api_key(action_id)
     extra_env_vars = {"MATRICE_INTERNAL_API_KEY": internal_api_key}
-    container_name = f"image_build_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(mount_docker_sock=True, extra_env_vars=extra_env_vars, container_name=container_name)} python3 main.py {model_family_id} {action_id}"'
+    cmd = f'{self.get_base_docker_cmd(mount_docker_sock=True, extra_env_vars=extra_env_vars)} python3 main.py {model_family_id} {action_id}"'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "image_build_log")
 
@@ -1999,8 +1740,7 @@ def resource_clone_execute(self: ActionInstance):
     if not action_details:
         return
     self.setup_action_requirements(action_details)
-    container_name = f"resource_clone_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(container_name=container_name)} python3 main.py {self.action_record_id} "'
+    cmd = f'{self.get_base_docker_cmd()} python3 main.py {self.action_record_id} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "resource_clone")
 
@@ -2016,27 +1756,17 @@ def streaming_gateway_execute(self: ActionInstance):
         self.docker_container = (
             f"aiforeveryone/streaming-gateway:{os.environ.get('ENV', 'prod')}"
         )
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for streaming gateway: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "streaming_gateway")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
-
-    container_name = f"streaming_gateway_{self.action_record_id}"
-    cmd = f'{self.get_base_docker_cmd(extra_pkgs=["matrice_streaming"], container_name=container_name)} python3 /usr/src/app/streaming_gateway.py {self.action_record_id} "'
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for training: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "streaming_gateway")
+        return
+    
+    cmd = f'{self.get_base_docker_cmd(extra_pkgs=["matrice_streaming"])} python3 /usr/src/app/streaming_gateway.py {self.action_record_id} "'
     logging.info("cmd is: %s", cmd)
     self.start(cmd, "streaming_gateway")
 
@@ -2130,24 +1860,16 @@ def kafka_setup_execute(self: ActionInstance):
     else:
         pkgs = f"matrice_common matrice"
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for kafka: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "kafka_setup")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for training: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "kafka_setup")
+        return
+
 
     # Kafka container with --net=host (Ports: 9092, 9093)
     cmd = (
@@ -2184,31 +1906,21 @@ def inference_tracker_setup_execute(self: ActionInstance):
     
     self.setup_action_requirements(action_details)
 
-    existing_container_id = action_details["actionDetails"].get("containerId")
-    if existing_container_id:
-        # Check if container actually exists before trying to restart
-        if ActionInstance.container_exists(existing_container_id):
-            logging.info(
-                "Using existing container ID for inference tracker: %s",
-                existing_container_id,
-            )
-            self.docker_container = existing_container_id
-            cmd = "docker restart " + self.docker_container
-            self.start(cmd, "inference_tracker_setup")
-            return
-        else:
-            logging.warning(
-                "Container %s not found. Creating new container.",
-                existing_container_id
-            )
-            # Fall through to create new container
-
+    if action_details["actionDetails"].get("containerId"):
+        logging.info(
+            "Using existing container ID for inference tracker: %s",
+            action_details["actionDetails"]["containerId"],
+        )
+        self.docker_container = action_details["actionDetails"]["containerId"]
+        cmd = "docker restart " + self.docker_container 
+        self.start(cmd, "inference_tracker_setup")
+        return
+    
     # This is the existing Docker run command
-    container_name = f"inference_tracker_{self.action_record_id}"
     worker_cmd = (
         f"docker run -d --pull=always --net=host "
-        f"--cidfile ./{self.action_record_id}.cid "
-        f"--name {container_name} "
+         f"--cidfile ./{self.action_record_id}.cid "
+        f"--name inference-tracker-worker "
         f"-v matrice_myvol:/matrice_data "
         f'-e ENV="{os.environ.get("ENV", "prod")}" '
         f'-e MATRICE_SECRET_ACCESS_KEY="{self.matrice_secret_access_key}" '
@@ -2256,7 +1968,7 @@ def video_storage_setup_execute(self: ActionInstance):
         f'-e MATRICE_SECRET_ACCESS_KEY="{self.matrice_secret_access_key}" '
         f'-e MATRICE_ACCESS_KEY_ID="{self.matrice_access_key_id}" '
         f'-e ACTION_ID="{self.action_record_id}" '
-        f'--restart=unless-stopped '
+        f' --restart=unless-stopped '
         f"{image}"
     )
     

{matrice_compute-0.1.35.dist-info → matrice_compute-0.1.36.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: matrice_compute
-Version: 0.1.35
+Version: 0.1.36
 Summary: Common server utilities for Matrice.ai services
 Author-email: "Matrice.ai" <dipendra@matrice.ai>
 License-Expression: MIT

{matrice_compute-0.1.35.dist-info → matrice_compute-0.1.36.dist-info}/RECORD

@@ -1,5 +1,5 @@
 matrice_compute/__init__.py,sha256=YZhx7rQlD1TAlhBMbsU3_Xp-tpLyTAxWZDcQvqmwR2g,723
-matrice_compute/action_instance.py,sha256=zVsW7-O3u5QoeZlh2D9Qlo5HsQNMOrUGbaR1jY1QFFg,88286
+matrice_compute/action_instance.py,sha256=03TX2dF2i2DUtMJvFJFckzvIEPsuyaJuNk9mkHWjsLM,75901
 matrice_compute/actions_manager.py,sha256=a_TulMnu462xc0t_A-Mpug5zhQTmtpjiv7mhiC_IAVw,18280
 matrice_compute/actions_scaledown_manager.py,sha256=pJ0nduNwHWZ10GnqJNx0Ok7cVWabQ_M8E2Vb9pH3A_k,2002
 matrice_compute/compute_operations_handler.py,sha256=amcMhmXtv2irE6qK8Vbgec_8uFqjWmVVp0VWq-73_MU,17781
@@ -11,8 +11,8 @@ matrice_compute/resources_tracker.py,sha256=DffKitGU1gran0OAuKIsfH0XeOe03xU7NGl-
 matrice_compute/scaling.py,sha256=UQDI8wN9JEKafvUVPF0Pk9XmhKlbMkeu16AZyyOuSE8,55147
 matrice_compute/shutdown_manager.py,sha256=rnP9Qes6JJKDnebmBC9rqkH__X9a8TMjhWQPWoOQKFs,13232
 matrice_compute/task_utils.py,sha256=3qIutiQdYPyGRxH9ZwLbqdg8sZcnp6jp08pszWCRFl0,2820
-matrice_compute-0.1.35.dist-info/licenses/LICENSE.txt,sha256=_uQUZpgO0mRYL5-fPoEvLSbNnLPv6OmbeEDCHXhK6Qc,1066
-matrice_compute-0.1.35.dist-info/METADATA,sha256=iP3k-FJhne2q4kbSSZx6IcTX1MPDr5upMcFeG0I_z6g,1038
-matrice_compute-0.1.35.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-matrice_compute-0.1.35.dist-info/top_level.txt,sha256=63Plr3L1GzBUWZO5JZaFkiv8IcB10xUPU-9w3i6ptvE,16
-matrice_compute-0.1.35.dist-info/RECORD,,
+matrice_compute-0.1.36.dist-info/licenses/LICENSE.txt,sha256=_uQUZpgO0mRYL5-fPoEvLSbNnLPv6OmbeEDCHXhK6Qc,1066
+matrice_compute-0.1.36.dist-info/METADATA,sha256=S3V1TndESfRIbXxF4M6CMxoqNtVZ3uvnoa7WkzcSNxI,1038
+matrice_compute-0.1.36.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+matrice_compute-0.1.36.dist-info/top_level.txt,sha256=63Plr3L1GzBUWZO5JZaFkiv8IcB10xUPU-9w3i6ptvE,16
+matrice_compute-0.1.36.dist-info/RECORD,,