Poetry: use locked environment in Docker images (#12385)

.dockerignore

@@ -4,8 +4,12 @@
 # things to include
 !docker
 !synapse
-!MANIFEST.in
 !README.rst
+!pyproject.toml
+!poetry.lock
+
+# TODO: remove these once we have moved over to using poetry-core in pyproject.toml
+!MANIFEST.in
 !setup.py
 
 **/__pycache__

changelog.d/12385.docker

@@ -0,0 +1 @@
+Bundle locked versions of dependencies into the Docker image.

docker/Dockerfile

@@ -14,20 +14,61 @@
 #    DOCKER_BUILDKIT=1 docker build -f docker/Dockerfile --build-arg PYTHON_VERSION=3.10 .
 #
 
+# Irritatingly, there is no blessed guide on how to distribute an application with its
+# poetry-managed environment in a docker image. We have opted for
+# `poetry export | pip install -r /dev/stdin`, but there are known bugs in
+# in `poetry export` whose fixes (scheduled for poetry 1.2) have yet to be released.
+# In case we get bitten by those bugs in the future, the recommendations here might
+# be useful:
+#     https://github.com/python-poetry/poetry/discussions/1879#discussioncomment-216865
+#     https://stackoverflow.com/questions/53835198/integrating-python-poetry-with-docker?answertab=scoredesc
+
+
+
 ARG PYTHON_VERSION=3.9
 
 ###
-### Stage 0: builder
+### Stage 0: generate requirements.txt
+###
+FROM docker.io/python:${PYTHON_VERSION}-slim as requirements
+
+# RUN --mount is specific to buildkit and is documented at
+# https://github.com/moby/buildkit/blob/master/frontend/dockerfile/docs/syntax.md#build-mounts-run---mount.
+# Here we use it to set up a cache for apt (and below for pip), to improve
+# rebuild speeds on slow connections.
+RUN \
+   --mount=type=cache,target=/var/cache/apt,sharing=locked \
+   --mount=type=cache,target=/var/lib/apt,sharing=locked \
+ apt-get update && apt-get install -y git \
+    && rm -rf /var/lib/apt/lists/*
+
+# We install poetry in its own build stage to avoid its dependencies conflicting with
+# synapse's dependencies.
+# We use a specific commit from poetry's master branch instead of our usual 1.1.12,
+# to incorporate fixes to some bugs in `poetry export`. This commit corresponds to
+#    https://github.com/python-poetry/poetry/pull/5156 and
+#    https://github.com/python-poetry/poetry/issues/5141 ;
+# without it, we generate a requirements.txt with incorrect environment markers,
+# which causes necessary packages to be omitted when we `pip install`.
+#
+# NB: In poetry 1.2 `poetry export` will be moved into a plugin; we'll need to also
+# pip install poetry-plugin-export (https://github.com/python-poetry/poetry-plugin-export).
+RUN --mount=type=cache,target=/root/.cache/pip \
+  pip install --user git+https://github.com/python-poetry/poetry.git@fb13b3a676f476177f7937ffa480ee5cff9a90a5
+
+WORKDIR /synapse
+
+# Copy just what we need to run `poetry export`...
+COPY pyproject.toml poetry.lock README.rst /synapse/
+
+RUN /root/.local/bin/poetry export --extras all -o /synapse/requirements.txt
+
+###
+### Stage 1: builder
 ###
 FROM docker.io/python:${PYTHON_VERSION}-slim as builder
 
 # install the OS build deps
-#
-# RUN --mount is specific to buildkit and is documented at
-# https://github.com/moby/buildkit/blob/master/frontend/dockerfile/docs/syntax.md#build-mounts-run---mount.
-# Here we use it to set up a cache for apt, to improve rebuild speeds on
-# slow connections.
-#
 RUN \
    --mount=type=cache,target=/var/cache/apt,sharing=locked \
    --mount=type=cache,target=/var/lib/apt,sharing=locked \
@@ -45,30 +86,27 @@ RUN \
     zlib1g-dev \
     && rm -rf /var/lib/apt/lists/*
 
-# Copy just what we need to pip install
-COPY MANIFEST.in README.rst setup.py /synapse/
-COPY synapse/__init__.py /synapse/synapse/__init__.py
-COPY synapse/python_dependencies.py /synapse/synapse/python_dependencies.py
-
 # To speed up rebuilds, install all of the dependencies before we copy over
-# the whole synapse project so that we this layer in the Docker cache can be
+# the whole synapse project, so that this layer in the Docker cache can be
 # used while you develop on the source
 #
-# This is aiming at installing the `install_requires` and `extras_require` from `setup.py`
+# This is aiming at installing the `[tool.poetry.depdendencies]` from pyproject.toml.
+COPY --from=requirements /synapse/requirements.txt /synapse/
 RUN --mount=type=cache,target=/root/.cache/pip \
-  pip install --prefix="/install" --no-warn-script-location \
-    /synapse[all]
+  pip install --prefix="/install" --no-warn-script-location -r /synapse/requirements.txt
 
-# Copy over the rest of the project
+# Copy over the rest of the synapse source code.
 COPY synapse /synapse/synapse/
+# ... and what we need to `pip install`.
+# TODO: once pyproject.toml declares poetry-core as its build system, we'll need to copy
+# pyproject.toml here, ditching setup.py and MANIFEST.in.
+COPY setup.py MANIFEST.in README.rst /synapse/
 
-# Install the synapse package itself and all of its children packages.
-#
-# This is aiming at installing only the `packages=find_packages(...)` from `setup.py
+# Install the synapse package itself.
 RUN pip install --prefix="/install" --no-deps --no-warn-script-location /synapse
 
 ###
-### Stage 1: runtime
+### Stage 2: runtime
 ###
 
 FROM docker.io/python:${PYTHON_VERSION}-slim

docker/start.py

@@ -108,7 +108,7 @@ def generate_config_from_template(config_dir, config_path, environ, ownership):
 
     # Hopefully we already have a signing key, but generate one if not.
     args = [
-        "python",
+        sys.executable,
         "-m",
         "synapse.app.homeserver",
         "--config-path",
@@ -158,7 +158,7 @@ def run_generate_config(environ, ownership):
 
     # generate the main config file, and a signing key.
     args = [
-        "python",
+        sys.executable,
         "-m",
         "synapse.app.homeserver",
         "--server-name",
@@ -175,7 +175,7 @@ def run_generate_config(environ, ownership):
         "--open-private-ports",
     ]
     # log("running %s" % (args, ))
-    os.execv("/usr/local/bin/python", args)
+    os.execv(sys.executable, args)
 
 
 def main(args, environ):
@@ -254,12 +254,12 @@ running with 'migrate_config'. See the README for more details.
 
     log("Starting synapse with args " + " ".join(args))
 
-    args = ["python"] + args
+    args = [sys.executable] + args
     if ownership is not None:
         args = ["gosu", ownership] + args
         os.execve("/usr/sbin/gosu", args, environ)
     else:
-        os.execve("/usr/local/bin/python", args, environ)
+        os.execve(sys.executable, args, environ)
 
 
 if __name__ == "__main__":