Hot fix v0.2.0

CHANGELOG.md

@@ -1,5 +1,10 @@
 # Changelog
 
+## [0.2.1]
+- Misc hot fixes
+- `LocalDriver` pull image if it doesn't exist locally.
+- Fix Compute Manager healthcheck
+
 ## [0.2.0]
 - The objective of this release is to support mounting the Flint Metastore as a POSIX-like filesystem. Consequently, reducing the number of Docker volumes required by the Control Plane.
 - The Experiment Tracker has been collapsed into the Experiment Server for simplicity and improved robustness around `inotify` events.

README.md

@@ -3,7 +3,7 @@
   <img width="60%" src="docs/_assets/logo-text.png" alt="FlintML Logo Text" /><br/>
 
   <!-- Badges, all inside the same HTML block -->
-  <img src="https://img.shields.io/badge/version-v0.2.0-cf051c" alt="Version 0.2.0" />
+  <img src="https://img.shields.io/badge/version-v0.2.1-cf051c" alt="Version 0.2.1" />
   <img src="https://img.shields.io/badge/license-BSL_1.1-blue" alt="License BSL 1.1" />
 
   </br>

VERSION

@@ -1 +1 @@
-0.2.0
+0.2.1

src/compute-manager/Dockerfile

@@ -12,6 +12,11 @@ RUN poetry config virtualenvs.create false && \
 
 FROM python:3.12-slim AS runtime
 
+# install curl
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends curl && \
+    rm -rf /var/lib/apt/lists/*
+
 WORKDIR /app
 
 # Inject dependencies

src/compute-manager/src/driver/local.py

@@ -1,7 +1,7 @@
 import asyncio
 import docker
 from docker.models.containers import Container
-from docker.errors import NotFound, APIError
+from docker.errors import ImageNotFound, APIError
 from docker.types import LogConfig
 from typing import Dict
 import os
@@ -25,9 +25,31 @@ def __init__(self, config: dict):
         super().__init__(config)
         self._docker: docker.DockerClient = docker.from_env()
 
+        self._ensure_worker_image()
+
         self._containers: Dict[str, Tuple[ContainerContext, Optional[Container]]] = {}
         self._watch_tasks: Dict[str, asyncio.Task] = {}
 
+    def _ensure_worker_image(self) -> None:
+        """
+        Check for self.worker_image in the local cache, and pull from Docker Hub
+        if it's not found.
+        """
+        try:
+            self._docker.images.get(self.worker_image)
+            logging.info(f"Image {self.worker_image} already present, skipping pull.")
+        except ImageNotFound:
+            logging.info(f"Image {self.worker_image} not found locally, pulling…")
+            try:
+                self._docker.images.pull(self.worker_image)
+                logging.info(f"Successfully pulled {self.worker_image}.")
+            except APIError as e:
+                logging.error(f"Failed to pull {self.worker_image}: {e}")
+                raise e
+        except APIError as e:
+            logging.error(f"Docker error inspecting image {self.worker_image}: {e}")
+            raise e
+
     async def launch_container(self, ctx: ContainerContext) -> None:
         """Start a container and begin watching it for unexpected exits."""
         try:

src/docker-compose.build.yml

@@ -61,7 +61,12 @@ services:
       context: .
       dockerfile: ./workspace/Dockerfile
     depends_on:
-      - catalog-explorer
+      storage:
+        condition: service_healthy
+      compute-manager:
+        condition: service_healthy
+      catalog-explorer:
+        condition: service_started
     restart: always
 
   reverse-proxy:

src/docker-compose.release-template.yml

@@ -55,7 +55,12 @@ services:
     <<: [*storage-creds, *s3fs]
     image: flintml/workspace:${VERSION}
     depends_on:
-      - catalog-explorer
+      storage:
+        condition: service_healthy
+      compute-manager:
+        condition: service_healthy
+      catalog-explorer:
+        condition: service_started
     restart: always
 
   reverse-proxy:

src/experiment-server/entrypoint.sh

@@ -39,5 +39,4 @@ server_pid=$!
 aim up \
     --repo /mnt/metastore/experiment \
     --host 0.0.0.0 --port 43800 \
-    --base-path /experiment-tracker \
-    --log-level debug
+    --base-path /experiment-tracker