🎨 create efs guardian specific user (#5936)

Author: matusdrobuliak66

.env-devel

@@ -65,6 +65,10 @@ DIRECTOR_PORT=8080
 DIRECTOR_REGISTRY_CACHING_TTL=900
 DIRECTOR_REGISTRY_CACHING=True
 
+EFS_USER_ID=8006
+EFS_USER_NAME=efs
+EFS_GROUP_ID=8106
+EFS_GROUP_NAME=efs-group
 EFS_DNS_NAME=fs-xxx.efs.us-east-1.amazonaws.com
 EFS_MOUNTED_PATH=/tmp/efs
 EFS_PROJECT_SPECIFIC_DATA_DIRECTORY=project-specific-data
@@ -169,6 +173,10 @@ R_CLONE_OPTION_RETRIES=3
 R_CLONE_OPTION_TRANSFERS=5
 R_CLONE_PROVIDER=MINIO
 
+# simcore-user used in docker images
+SC_USER_ID=8004
+SC_USER_NAME=scu
+
 S3_ACCESS_KEY=12345678
 S3_BUCKET_NAME=simcore
 S3_ENDPOINT=http://172.17.0.1:9001

services/docker-compose.yml

@@ -365,6 +365,12 @@ services:
       RABBIT_PORT: ${RABBIT_PORT}
       RABBIT_SECURE: ${RABBIT_SECURE}
       RABBIT_USER: ${RABBIT_USER}
+      SC_USER_ID: ${SC_USER_ID}
+      SC_USER_NAME: ${SC_USER_NAME}
+      EFS_USER_ID: ${EFS_USER_ID}
+      EFS_USER_NAME: ${EFS_USER_NAME}
+      EFS_GROUP_ID: ${EFS_GROUP_ID}
+      EFS_GROUP_NAME: ${EFS_GROUP_NAME}
       EFS_DNS_NAME: ${EFS_DNS_NAME}
       EFS_MOUNTED_PATH: ${EFS_MOUNTED_PATH}
       EFS_ONLY_ENABLED_FOR_USERIDS: ${EFS_ONLY_ENABLED_FOR_USERIDS}

services/efs-guardian/Dockerfile

@@ -45,26 +45,44 @@ RUN --mount=type=cache,target=/var/cache/apt,mode=0755,sharing=private \
   && gosu nobody true
 
 # simcore-user uid=8004(scu) gid=8004(scu) groups=8004(scu)
+# simcore-efs-user uid=8006(efs) gid=8006(efs) groups=8006(efs)
+# Currently all simcore services run as user 8004. The Guardian needs to run as a different user 8006.
+# The Guardian has access to the root directory of the EFS distributed file system
+# and can change the file permissions of user 8004 if needed.
 ENV SC_USER_ID=8004 \
   SC_USER_NAME=scu \
+  EFS_USER_ID=8006 \
+  EFS_USER_NAME=efs \
   SC_BUILD_TARGET=base \
   SC_BOOT_MODE=default
 
 RUN adduser \
-  --uid ${SC_USER_ID} \
+  --uid 8004 \
   --disabled-password \
   --gecos "" \
   --shell /bin/sh \
-  --home /home/${SC_USER_NAME} \
-  ${SC_USER_NAME}
+  --home /home/scu \
+  scu
 
+RUN adduser \
+  --uid 8006 \
+  --disabled-password \
+  --gecos "" \
+  --shell /bin/sh \
+  --home /home/efs \
+  efs
+
+# we create efs-group which is then used in efs guardian when he is creating a directory for user services.
+RUN groupadd -g 8106 efs-group
+RUN usermod -a -G efs-group efs
+RUN usermod -a -G efs-group scu
 
 # Sets utf-8 encoding for Python et al
 ENV LANG=C.UTF-8
 
 # Turns off writing .pyc files; superfluous on an ephemeral container.
 ENV PYTHONDONTWRITEBYTECODE=1 \
-  VIRTUAL_ENV=/home/scu/.venv
+  VIRTUAL_ENV=/home/efs/.venv
 
 # Ensures that the python and pip executables used in the image will be
 # those from our virtualenv.
@@ -149,15 +167,15 @@ ENV SC_BUILD_TARGET=production \
 
 ENV PYTHONOPTIMIZE=TRUE
 
-WORKDIR /home/scu
-# ensure home folder is read/writable for user scu
-RUN chown -R scu /home/scu
+WORKDIR /home/efs
+# ensure home folder is read/writable for user efs
+RUN chown -R efs /home/efs
 
 # Starting from clean base image, copies pre-installed virtualenv from prod-only-deps
-COPY --chown=scu:scu --from=prod-only-deps  ${VIRTUAL_ENV} ${VIRTUAL_ENV}
+COPY --chown=efs:efs --from=prod-only-deps  ${VIRTUAL_ENV} ${VIRTUAL_ENV}
 
 # Copies booting scripts
-COPY --chown=scu:scu services/efs-guardian/docker services/efs-guardian/docker
+COPY --chown=efs:efs services/efs-guardian/docker services/efs-guardian/docker
 RUN chmod +x services/efs-guardian/docker/*.sh
 
 
@@ -187,7 +205,7 @@ ENV SC_BUILD_TARGET=development \
 
 WORKDIR /devel
 
-RUN chown -R scu:scu "${VIRTUAL_ENV}"
+RUN chown -R efs:efs "${VIRTUAL_ENV}"
 
 ENTRYPOINT ["/bin/sh", "services/efs-guardian/docker/entrypoint.sh"]
 CMD ["/bin/sh", "services/efs-guardian/docker/boot.sh"]

services/efs-guardian/docker/entrypoint.sh

@@ -43,30 +43,30 @@ if [ "${SC_BUILD_TARGET}" = "development" ]; then
   HOST_GROUPID=$(stat --format=%g "${SC_DEVEL_MOUNT}")
   CONT_GROUPNAME=$(getent group "${HOST_GROUPID}" | cut --delimiter=: --fields=1)
   if [ "$HOST_USERID" -eq 0 ]; then
-    echo "$WARNING" "Folder mounted owned by root user... adding $SC_USER_NAME to root..."
-    adduser "$SC_USER_NAME" root
+    echo "$WARNING" "Folder mounted owned by root user... adding $EFS_USER_NAME to root..."
+    adduser "$EFS_USER_NAME" root
   else
     echo "$INFO" "Folder mounted owned by user $HOST_USERID:$HOST_GROUPID-'$CONT_GROUPNAME'..."
-    # take host's credentials in $SC_USER_NAME
+    # take host's credentials in $EFS_USER_NAME
     if [ -z "$CONT_GROUPNAME" ]; then
-      echo "$WARNING" "Creating new group grp$SC_USER_NAME"
-      CONT_GROUPNAME=grp$SC_USER_NAME
+      echo "$WARNING" "Creating new group grp$EFS_USER_NAME"
+      CONT_GROUPNAME=grp$EFS_USER_NAME
       addgroup --gid "$HOST_GROUPID" "$CONT_GROUPNAME"
     else
       echo "$INFO" "group already exists"
     fi
-    echo "$INFO" "Adding $SC_USER_NAME to group $CONT_GROUPNAME..."
-    adduser "$SC_USER_NAME" "$CONT_GROUPNAME"
+    echo "$INFO" "Adding $EFS_USER_NAME to group $CONT_GROUPNAME..."
+    adduser "$EFS_USER_NAME" "$CONT_GROUPNAME"
 
     echo "$WARNING" "Changing ownership [this could take some time]"
-    echo "$INFO" "Changing $SC_USER_NAME:$SC_USER_NAME ($SC_USER_ID:$SC_USER_ID) to $SC_USER_NAME:$CONT_GROUPNAME ($HOST_USERID:$HOST_GROUPID)"
-    usermod --uid "$HOST_USERID" --gid "$HOST_GROUPID" "$SC_USER_NAME"
+    echo "$INFO" "Changing $EFS_USER_NAME:$EFS_USER_NAME ($EFS_USER_ID:$EFS_USER_ID) to $EFS_USER_NAME:$CONT_GROUPNAME ($HOST_USERID:$HOST_GROUPID)"
+    usermod --uid "$HOST_USERID" --gid "$HOST_GROUPID" "$EFS_USER_NAME"
 
-    echo "$INFO" "Changing group properties of files around from $SC_USER_ID to group $CONT_GROUPNAME"
-    find / -path /proc -prune -o -group "$SC_USER_ID" -exec chgrp --no-dereference "$CONT_GROUPNAME" {} \;
+    echo "$INFO" "Changing group properties of files around from $EFS_USER_ID to group $CONT_GROUPNAME"
+    find / -path /proc -prune -o -group "$EFS_USER_ID" -exec chgrp --no-dereference "$CONT_GROUPNAME" {} \;
     # change user property of files already around
-    echo "$INFO" "Changing ownership properties of files around from $SC_USER_ID to group $CONT_GROUPNAME"
-    find / -path /proc -prune -o -user "$SC_USER_ID" -exec chown --no-dereference "$SC_USER_NAME" {} \;
+    echo "$INFO" "Changing ownership properties of files around from $EFS_USER_ID to group $CONT_GROUPNAME"
+    find / -path /proc -prune -o -user "$EFS_USER_ID" -exec chown --no-dereference "$EFS_USER_NAME" {} \;
   fi
 fi
 
@@ -84,11 +84,11 @@ if stat $DOCKER_MOUNT >/dev/null 2>&1; then
     GROUPNAME=$(getent group "${GROUPID}" | cut --delimiter=: --fields=1)
     echo "$WARNING docker group with $GROUPID has name $GROUPNAME"
   fi
-  adduser "$SC_USER_NAME" "$GROUPNAME"
+  adduser "$EFS_USER_NAME" "$GROUPNAME"
 fi
 
 echo "$INFO Starting $* ..."
-echo "  $SC_USER_NAME rights    : $(id "$SC_USER_NAME")"
+echo "  $EFS_USER_NAME rights    : $(id "$EFS_USER_NAME")"
 echo "  local dir : $(ls -al)"
 
-exec gosu "$SC_USER_NAME" "$@"
+exec gosu "$EFS_USER_NAME" "$@"

services/efs-guardian/src/simcore_service_efs_guardian/core/settings.py

@@ -41,8 +41,21 @@ class ApplicationSettings(BaseCustomSettings, MixinLoggingSettings):
         "then the check is considered to have failed."
         "It takes retries consecutive failures of the health check for the container to be considered unhealthy.",
     )
-    SC_USER_ID: int | None = None
-    SC_USER_NAME: str | None = None
+    SC_USER_ID: int
+    SC_USER_NAME: str
+
+    EFS_USER_ID: int = Field(
+        description="Linux user ID that the Guardian service will run with"
+    )
+    EFS_USER_NAME: str = Field(
+        description="Linux user name that the Guardian service will run with"
+    )
+    EFS_GROUP_ID: int = Field(
+        description="Linux group ID that the EFS and Simcore linux users are part of"
+    )
+    EFS_GROUP_NAME: str = Field(
+        description="Linux group name that the EFS and Simcore linux users are part of"
+    )
 
     # RUNTIME  -----------------------------------------------------------
     EFS_GUARDIAN_DEBUG: bool = Field(

services/efs-guardian/src/simcore_service_efs_guardian/services/efs_manager.py

@@ -1,17 +1,21 @@
+import os
 from dataclasses import dataclass
 from pathlib import Path
 
 from fastapi import FastAPI
 from models_library.projects import ProjectID
 from models_library.projects_nodes_io import NodeID
 
+from ..core.settings import ApplicationSettings, get_application_settings
+
 
 @dataclass(frozen=True)
 class EfsManager:
     app: FastAPI
 
     _efs_mounted_path: Path
     _project_specific_data_base_directory: str
+    _settings: ApplicationSettings
 
     @classmethod
     async def create(
@@ -20,7 +24,10 @@ async def create(
         efs_mounted_path: Path,
         project_specific_data_base_directory: str,
     ):
-        return cls(app, efs_mounted_path, project_specific_data_base_directory)
+        settings = get_application_settings(app)
+        return cls(
+            app, efs_mounted_path, project_specific_data_base_directory, settings
+        )
 
     async def initialize_directories(self):
         _dir_path = self._efs_mounted_path / self._project_specific_data_base_directory
@@ -36,5 +43,12 @@ async def create_project_specific_data_dir(
             / f"{node_id}"
             / f"{storage_directory_name}"
         )
+        # Ensure the directory exists with the right parents
         Path.mkdir(_dir_path, parents=True, exist_ok=True)
+        # Change the owner to user id 8006(efs) and group id 8106(efs-group)
+        os.chown(_dir_path, self._settings.EFS_USER_ID, self._settings.EFS_GROUP_ID)
+        # Set directory permissions to allow group write access
+        Path.chmod(
+            _dir_path, 0o770
+        )  # This gives rwx permissions to user and group, and nothing to others
         return _dir_path

services/efs-guardian/tests/unit/test_efs_manager.py

@@ -5,6 +5,7 @@
 # pylint: disable=unused-variable
 
 from pathlib import Path
+from unittest.mock import patch
 
 import pytest
 from faker import Faker
@@ -45,12 +46,17 @@ async def test_rpc_create_project_specific_data_dir(
     _node_id = faker.uuid4()
     _storage_directory_name = faker.name()
 
-    result = await efs_manager.create_project_specific_data_dir(
-        rpc_client,
-        project_id=_project_id,
-        node_id=_node_id,
-        storage_directory_name=_storage_directory_name,
-    )
+    with patch(
+        "simcore_service_efs_guardian.services.efs_manager.os.chown"
+    ) as mocked_chown:
+        result = await efs_manager.create_project_specific_data_dir(
+            rpc_client,
+            project_id=_project_id,
+            node_id=_node_id,
+            storage_directory_name=_storage_directory_name,
+        )
+        mocked_chown.assert_called_once()
+
     assert isinstance(result, Path)
     _expected_path = (
         aws_efs_settings.EFS_MOUNTED_PATH