matrice-compute 0.1.30__tar.gz → 0.1.31__tar.gz

{matrice_compute-0.1.30 → matrice_compute-0.1.31}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: matrice_compute
-Version: 0.1.30
+Version: 0.1.31
 Summary: Common server utilities for Matrice.ai services
 Author-email: "Matrice.ai" <dipendra@matrice.ai>
 License-Expression: MIT

{matrice_compute-0.1.30 → matrice_compute-0.1.31}/matrice_compute.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: matrice_compute
-Version: 0.1.30
+Version: 0.1.31
 Summary: Common server utilities for Matrice.ai services
 Author-email: "Matrice.ai" <dipendra@matrice.ai>
 License-Expression: MIT

{matrice_compute-0.1.30 → matrice_compute-0.1.31}/src/matrice_compute/action_instance.py

@@ -490,12 +490,11 @@ class ActionInstance:
         )
 
         cmd_parts = [
-            f"docker run {use_gpu} ",
+            f"docker run -d {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} && "
@@ -897,60 +896,45 @@ class ActionInstance:
         self.cmd = cmd
         self.log_path = f"{self.get_log_path()}/{log_name}_{self.action_record_id}.txt"
 
-        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,
-            )
+        # 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},
+        )
         
-        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:
+        stdout, stderr = process.communicate(timeout=120)
+        
+        if process.returncode != 0:
             logging.error(
-                "Failed to read CID file for action %s after %d attempts",
+                "Docker run failed for action %s: %s",
                 self.action_record_id,
-                max_retries,
+                stderr,
             )
-            raise Exception("Failed to start process: CID file not found")
+            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,
+        )
         
-        # report container id to scaling service
+        # 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,
+        )
+        
+        # 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):