Feature: Plugin V2: Running plugin on host (#461)

Author: timotheeguerin

.vscode/settings.json

@@ -14,5 +14,7 @@
   "python.formatting.provider": "yapf",
   "python.formatting.yapfArgs": [
     "--style=.style.yapf"
-  ]
+  ],
+  "python.venvPath": "${workspaceFolder}/ENV",
+  "python.pythonPath": "${workspaceFolder}\\ENV\\Scripts\\python.exe"
 }

aztk/error.py

@@ -6,7 +6,7 @@
 
 
 class AztkError(Exception):
-    def __init__(self, message: str = None):
+    def __init__(self, message: str=None):
         super().__init__(message)
 
 class ClusterNotReadyError(AztkError):

aztk/internal/__init__.py

@@ -3,3 +3,4 @@
 """
 
 from .configuration_base import *
+from .docker_cmd import *

aztk/internal/cluster_data/node_data.py

@@ -1,11 +1,10 @@
 import fnmatch
 import io
-import json
 import os
-import yaml
 import zipfile
 from pathlib import Path
 from typing import List
+import yaml
 from aztk.spark import models
 from aztk.utils import constants, file_utils, secure_utils
 from aztk.error import InvalidCustomScriptError
@@ -147,7 +146,8 @@ def _add_plugins(self):
                     execute='{0}/{1}'.format(plugin.name, plugin.execute),
                     args=plugin.args,
                     env=plugin.env,
-                    runOn=plugin.run_on.value,
+                    target=plugin.target.value,
+                    target_role=plugin.target_role.value,
                 ))
 
         self.zipf.writestr(os.path.join('plugins', 'plugins-manifest.yaml'), yaml.dump(data))

aztk/internal/docker_cmd.py

@@ -0,0 +1,38 @@
+import os
+from aztk.utils.command_builder import CommandBuilder
+
+class DockerCmd:
+    """
+    Class helping to write a docker command
+    """
+
+    def __init__(self, name: str, docker_repo: str, cmd: str, gpu_enabled=False):
+        if gpu_enabled:
+            self.cmd = CommandBuilder('nvidia-docker run')
+        else:
+            self.cmd = CommandBuilder('docker run')
+        self.cmd.add_option('--net', 'host')
+        self.cmd.add_option('--name', name)
+        self.cmd.add_argument('-d')
+        self.cmd.add_argument(docker_repo)
+        self.cmd.add_argument(cmd)
+
+
+    def add_env(self, env: str, value: str):
+        self.cmd.add_option('-e', '{0}={1}'.format(env, value))
+
+    def pass_env(self, env: str):
+        """
+        Give the value of an environment variable in the main process to the docker image
+        """
+        self.cmd.add_option('-e', '{0}'.format(env))
+
+    def share_folder(self, folder: str):
+        self.cmd.add_option('-v', '{0}:{0}'.format(folder))
+
+    def open_port(self, port: int):
+        self.cmd.add_option('-p', '{0}:{0}'.format(port))       # Spark Master UI
+
+
+    def to_str(self):
+        return self.cmd.to_str()

aztk/models/plugins/internal/plugin_manager.py

@@ -20,6 +20,7 @@ class PluginManager:
         jupyterlab=plugins.JupyterLabPlugin,
         rstudio_server=plugins.RStudioServerPlugin,
         hdfs=plugins.HDFSPlugin,
+        simple=plugins.SimplePlugin,
         spark_ui_proxy=plugins.SparkUIProxyPlugin,
     )
 

aztk/models/plugins/plugin_configuration.py

@@ -1,8 +1,23 @@
-import inspect
-from typing import List, Union
 from enum import Enum
-from .plugin_file import PluginFile
+from typing import List, Union
 from aztk.internal import ConfigurationBase
+from aztk.error import InvalidPluginConfigurationError
+from .plugin_file import PluginFile
+
+class PluginTarget(Enum):
+    """
+    Where this plugin should run
+    """
+    SparkContainer = "spark-container",
+    Host = "host",
+
+
+class PluginTargetRole(Enum):
+    Master = "master"
+    Worker = "worker"
+    All = "all-nodes"
+
+
 
 class PluginPort:
     """
@@ -12,8 +27,7 @@ class PluginPort:
         :param name: [Optional] name to differentiate ports if you have multiple
     """
 
-    def __init__(self, internal: int, public: Union[int, bool] = False, name=None):
-
+    def __init__(self, internal: int, public: Union[int, bool]=False, name=None):
         self.internal = internal
         self.expose_publicly = bool(public)
         self.public_port = None
@@ -26,11 +40,6 @@ def __init__(self, internal: int, public: Union[int, bool] = False, name=None):
         self.name = name
 
 
-class PluginRunTarget(Enum):
-    Master = "master"
-    Worker = "worker"
-    All = "all-nodes"
-
 
 
 class PluginConfiguration(ConfigurationBase):
@@ -45,15 +54,17 @@ class PluginConfiguration(ConfigurationBase):
 
     def __init__(self,
                  name: str,
-                 ports: List[PluginPort] = None,
-                 files: List[PluginFile] = None,
-                 execute: str = None,
+                 ports: List[PluginPort]=None,
+                 files: List[PluginFile]=None,
+                 execute: str=None,
                  args=None,
                  env=None,
-                 run_on: PluginRunTarget = PluginRunTarget.Master):
+                 target_role: PluginTargetRole=PluginTargetRole.Master,
+                 target: PluginTarget=PluginTarget.SparkContainer):
         self.name = name
         # self.docker_image = docker_image
-        self.run_on = run_on
+        self.target = target
+        self.target_role = target_role
         self.ports = ports or []
         self.files = files or []
         self.args = args or []
@@ -64,11 +75,18 @@ def has_arg(self, name: str):
         for x in self.args:
             if x.name == name:
                 return True
-        else:
-            return False
+        return False
 
     def validate(self):
         self._validate_required([
             "name",
             "execute",
         ])
+
+        if not isinstance(self.target, PluginTarget):
+            raise InvalidPluginConfigurationError(
+                "Target must be of type Plugin target but was {0}".format(self.target))
+
+        if not isinstance(self.target_role, PluginTargetRole):
+            raise InvalidPluginConfigurationError(
+                "Target role must be of type Plugin target role but was {0}".format(self.target))

aztk/node_scripts/core/config.py

@@ -16,14 +16,14 @@
                                  '/providers/[^/]+'
                                  '/[^/]+Accounts/(?P<account>[^/]+)$')
 
-batch_account_name = os.environ["AZ_BATCH_ACCOUNT_NAME"]
-batch_account_key = os.environ["BATCH_ACCOUNT_KEY"]
-batch_service_url = os.environ["BATCH_SERVICE_URL"]
-tenant_id = os.environ["SP_TENANT_ID"]
-client_id = os.environ["SP_CLIENT_ID"]
-credential = os.environ["SP_CREDENTIAL"]
-batch_resource_id = os.environ["SP_BATCH_RESOURCE_ID"]
-storage_resource_id = os.environ["SP_STORAGE_RESOURCE_ID"]
+batch_account_name = os.environ.get("AZ_BATCH_ACCOUNT_NAME")
+batch_account_key = os.environ.get("BATCH_ACCOUNT_KEY")
+batch_service_url = os.environ.get("BATCH_SERVICE_URL")
+tenant_id = os.environ.get("SP_TENANT_ID")
+client_id = os.environ.get("SP_CLIENT_ID")
+credential = os.environ.get("SP_CREDENTIAL")
+batch_resource_id = os.environ.get("SP_BATCH_RESOURCE_ID")
+storage_resource_id = os.environ.get("SP_STORAGE_RESOURCE_ID")
 
 pool_id = os.environ["AZ_BATCH_POOL_ID"]
 node_id = os.environ["AZ_BATCH_NODE_ID"]
@@ -33,9 +33,9 @@
 spark_worker_ui_port = os.environ["SPARK_WORKER_UI_PORT"]
 spark_job_ui_port = os.environ["SPARK_JOB_UI_PORT"]
 
-storage_account_name = os.environ["STORAGE_ACCOUNT_NAME"]
-storage_account_key = os.environ["STORAGE_ACCOUNT_KEY"]
-storage_account_suffix = os.environ["STORAGE_ACCOUNT_SUFFIX"]
+storage_account_name = os.environ.get("STORAGE_ACCOUNT_NAME")
+storage_account_key = os.environ.get("STORAGE_ACCOUNT_KEY")
+storage_account_suffix = os.environ.get("STORAGE_ACCOUNT_SUFFIX")
 
 def get_blob_client() -> blob.BlockBlobService:
     if not storage_resource_id:

aztk/node_scripts/docker_main.sh

@@ -3,12 +3,13 @@
 # This file is the entry point of the docker container.
 
 set -e
+echo "Initializing spark container"
 
 # --------------------
 # Setup custom scripts
 # --------------------
-custom_script_dir=$DOCKER_WORKING_DIR/custom-scripts
-aztk_dir=$DOCKER_WORKING_DIR/aztk
+custom_script_dir=$AZTK_WORKING_DIR/custom-scripts
+aztk_dir=$AZTK_WORKING_DIR/aztk
 
 # -----------------------
 # Preload jupyter samples
@@ -28,11 +29,11 @@ done
 echo "Starting setup using Docker"
 
 $(pyenv root)/versions/$AZTK_PYTHON_VERSION/bin/pip install -r $(dirname $0)/requirements.txt
-export PYTHONPATH=$PYTHONPATH:$DOCKER_WORKING_DIR
-echo 'export PYTHONPATH=$PYTHONPATH:$DOCKER_WORKING_DIR' >> ~/.bashrc
+export PYTHONPATH=$PYTHONPATH:$AZTK_WORKING_DIR
+echo 'export PYTHONPATH=$PYTHONPATH:$AZTK_WORKING_DIR' >> ~/.bashrc
 
 echo "Running main.py script"
-$(pyenv root)/versions/$AZTK_PYTHON_VERSION/bin/python $(dirname $0)/main.py install
+$(pyenv root)/versions/$AZTK_PYTHON_VERSION/bin/python $(dirname $0)/main.py setup-spark-container
 
 # sleep to keep container running
 while true; do sleep 1; done

aztk/node_scripts/install/create_user.py

@@ -6,11 +6,11 @@
 from datetime import datetime, timezone, timedelta
 import yaml
 '''
-    Creates a user if the user configuration file at $DOCKER_WORKING_DIR/user.yaml exists
+    Creates a user if the user configuration file at $AZTK_WORKING_DIR/user.yaml exists
 '''
 
 def create_user(batch_client):
-    path = os.path.join(os.environ['DOCKER_WORKING_DIR'], "user.yaml")
+    path = os.path.join(os.environ['AZTK_WORKING_DIR'], "user.yaml")
 
     if not os.path.isfile(path):
         print("No user to create.")
@@ -43,7 +43,7 @@ def decrypt_password(user_conf):
     tag = user_conf['tag']
 
     # Read private key
-    with open(os.path.join(os.environ['DOCKER_WORKING_DIR'], 'id_rsa'), encoding='UTF-8') as f:
+    with open(os.path.join(os.environ['AZTK_WORKING_DIR'], 'id_rsa'), encoding='UTF-8') as f:
         private_key = RSA.import_key(f.read())
     # Decrypt the session key with the public RSA key
     cipher_rsa = PKCS1_OAEP.new(private_key)