RE2022-217: Workspace uploader performance

src/loaders/common/callback_server_wrapper.py

@@ -16,70 +16,116 @@
 class Conf:
     """
     Configuration class for the workspace downloader and workspace uploader scripts.
+
+    Instance variables:
+
+    token - a KBase token appropriate for the KBase environment
+    callback_url - the url of the callback service to contact
+    job_data_dir - the directory for SDK jobs per user
+    input_queue - queue for the workspace downloader tasks
+    output_dir - the directory for a specific workspace id under sourcedata/ws
+    retrieve_sample - whether to retrieve sample for each genome object
+    ignore_no_sample_error - whether to ignore the error when no sample data is found
+    ws - workspace client
+    asu - assemblyUtil client
+    ss - sampleService client
+    pools - a pool of worker processes
+
     """
     def __init__(
-            self,
-            job_dir: str,
-            output_dir: str,
-            worker_function: Callable,
-            kb_base_url: str = "https://ci.kbase.us/services/",
-            token_filepath: str | None = None,
-            workers: int = 5,
-            retrieve_sample: bool = False,
-            ignore_no_sample_error: bool = False,
+        self,
+        job_dir: str,
+        output_dir: str | None = None,
+        kb_base_url: str = "https://ci.kbase.us/services/",
+        token_filepath: str | None = None,
+        au_service_ver: str = "release",
+        workers: int = 5,
+        max_callback_server_tasks: int = 20,
+        worker_function: Callable | None = None,
+        retrieve_sample: bool = False,
+        ignore_no_sample_error: bool = False,
+        workspace_downloader: bool = False,
+        catalog_admin: bool = False,
     ) -> None:
         """
         Initialize the configuration class.
 
         Args:
             job_dir (str): The directory for SDK jobs per user.
             output_dir (str): The directory for a specific workspace id under sourcedata/ws.
-            worker_function (Callable): The function that will be called by the workers.
             kb_base_url (str): The base url of the KBase services.
-            token_filepath (str): The file path that stores a KBase token appropriate for the KBase environment. 
+            token_filepath (str): The file path that stores a KBase token appropriate for the KBase environment.
                                 If not supplied, the token must be provided in the environment variable KB_AUTH_TOKEN.
+                                The KB_ADMIN_AUTH_TOKEN environment variable will get set by this token if the user runs as catalog admin.
+            au_service_ver (str): The service verison of AssemblyUtilClient
+                                ('dev', 'beta', 'release', or a git commit).
             workers (int): The number of workers to use for multiprocessing.
+            max_callback_server_tasks (int): The maxmium subtasks for the callback server.
+            worker_function (Callable): The function that will be called by the workers.
             retrieve_sample (bool): Whether to retrieve sample for each genome object.
             ignore_no_sample_error (bool): Whether to ignore the error when no sample data is found.
+            workspace_downloader (bool): Whether to be used for the workspace downloader script.
+            catalog_admin (bool): Whether to run the callback server as catalog admin.
         """
-        port = loader_helper.find_free_port()
-        token = loader_helper.get_token(token_filepath)
-        self.retrieve_sample = retrieve_sample
-        self.ignore_no_sample_error = ignore_no_sample_error
         ipv4 = loader_helper.get_ip()
+        port = loader_helper.find_free_port()
+
+        # common instance variables
+
+        self.token = loader_helper.get_token(token_filepath)
+
+        # setup and run callback server container
         self._start_callback_server(
             docker.from_env(),
             uuid.uuid4().hex,
             job_dir,
             kb_base_url,
-            token,
+            self.token,
             port,
+            max_callback_server_tasks,
             ipv4,
+            catalog_admin,
         )
 
-        ws_url = os.path.join(kb_base_url, "ws")
-        sample_url = os.path.join(kb_base_url, "sampleservice")
-        callback_url = "http://" + ipv4 + ":" + str(port)
-        print("callback_url:", callback_url)
-
-        self.ws = Workspace(ws_url, token=token)
-        self.asu = AssemblyUtil(callback_url, token=token)
-        self.ss = SampleService(sample_url, token=token)
+        self.callback_url = "http://" + ipv4 + ":" + str(port)
+        print("callback_url:", self.callback_url)
 
-        self.workers = workers
-        self.output_dir = output_dir
-        self.input_queue = Queue()
-        self.output_queue = Queue()
         self.job_data_dir = loader_helper.make_job_data_dir(job_dir)
-        self.pools = Pool(workers, worker_function, [self])
+
+        # unique to downloader
+        if workspace_downloader:
+            if worker_function is None:
+                raise ValueError(
+                    "worker_function cannot be None for the workspace downloader script"
+                )
+
+            self.input_queue = Queue()
+            self.output_dir = output_dir
+
+            self.retrieve_sample = retrieve_sample
+            self.ignore_no_sample_error = ignore_no_sample_error
+
+            ws_url = os.path.join(kb_base_url, "ws")
+            self.ws = Workspace(ws_url, token=self.token)
+
+            self.asu = AssemblyUtil(
+                self.callback_url, service_ver=au_service_ver, token=self.token
+            )
+
+            sample_url = os.path.join(kb_base_url, "sampleservice")
+            self.ss = SampleService(sample_url, token=self.token)
+
+            self.pools = Pool(workers, worker_function, [self])
 
     def _setup_callback_server_envs(
-            self,
-            job_dir: str,
-            kb_base_url: str,
-            token: str,
-            port: int,
-            ipv4: str,
+        self,
+        job_dir: str,
+        kb_base_url: str,
+        token: str,
+        port: int,
+        max_callback_server_tasks: int,
+        ipv4: str,
+        catalog_admin: bool,
     ) -> Tuple[dict[str, Union[int, str]], dict[str, dict[str, str]]]:
         """
         Setup the environment variables and volumes for the callback server.
@@ -89,6 +135,9 @@
             kb_base_url (str): The base url of the KBase services.
             token (str): The KBase token.
             port (int): The port number for the callback server.
+            max_callback_server_tasks (int): The maxmium subtasks for the callback server.
+            ipv4 (str): The ipv4 address for the callback server.
+            catalog_admin (bool): Whether to run the callback server as catalog admin.
 
         Returns:
             tuple: A tuple of the environment variables and volumes for the callback server.
@@ -102,9 +151,14 @@
         env["KB_BASE_URL"] = kb_base_url
         env["JOB_DIR"] = job_dir
         env["CALLBACK_PORT"] = port
+        env["JR_MAX_TASKS"] = max_callback_server_tasks
         env["CALLBACK_IP"] = ipv4  # specify an ipv4 address for the callback server
                                    # otherwise, the callback container will use the an ipv6 address
 
+        # set admin token to get catalog secure params
+        if catalog_admin:
+            env["KB_ADMIN_AUTH_TOKEN"] = token
+
         # setup volumes required for docker container
         docker_host = os.environ["DOCKER_HOST"]
         if docker_host.startswith("unix:"):
@@ -116,14 +170,16 @@
         return env, vol
 
     def _start_callback_server(
-            self,
-            client: docker.client,
-            container_name: str,
-            job_dir: str,
-            kb_base_url: str,
-            token: str,
-            port: int,
-            ipv4: str,
+        self,
+        client: docker.client,
+        container_name: str,
+        job_dir: str,
+        kb_base_url: str,
+        token: str,
+        port: int,
+        max_callback_server_tasks: int,
+        ipv4: str,
+        catalog_admin: bool,
     ) -> None:
         """
         Start the callback server.
@@ -134,9 +190,20 @@
             job_dir (str): The directory for SDK jobs per user.
             kb_base_url (str): The base url of the KBase services.
             token (str): The KBase token.
+            max_callback_server_tasks (int): The maxmium subtasks for the callback server.
             port (int): The port number for the callback server.
+            ipv4 (str): The ipv4 address for the callback server.
+            catalog_admin (bool): Whether to run the callback server as catalog admin.
         """
-        env, vol = self._setup_callback_server_envs(job_dir, kb_base_url, token, port, ipv4)
+        env, vol = self._setup_callback_server_envs(
+            job_dir,
+            kb_base_url,
+            token,
+            port,
+            max_callback_server_tasks,
+            ipv4,
+            catalog_admin,
+        )
         self.container = client.containers.run(
             name=container_name,
             image=CALLBACK_IMAGE_NAME,
@@ -147,9 +214,21 @@
         )
         time.sleep(2)
 
+    def _get_container_logs(self) -> None:
+        """
+        Get logs from the callback server container.
+        """
+        logs = self.container.logs()
+        if logs:
+            print("\n****** Logs from the Callback Server ******\n")
+            logs = logs.decode("utf-8")
+            for line in logs.split("\n"):
+                print(line)
+
     def stop_callback_server(self) -> None:
         """
         Stop the callback server.
         """
+        self._get_container_logs()
         self.container.stop()
         self.container.remove()

src/loaders/common/loader_common_names.py

@@ -129,6 +129,13 @@
                    'APPDEV': 'https://appdev.kbase.us/services/',
                    'PROD': 'https://kbase.us/services/'}
 
+# containers.conf path
+CONTAINERS_CONF_PATH = "~/.config/containers/containers.conf"
+# params in containers.conf file
+CONTAINERS_CONF_PARAMS = {
+    "seccomp_profile": "\"unconfined\"",
+    "log_driver": "\"k8s-file\""
+}
 # field name for Kbase object metadata
 FLD_KB_OBJ_UPA = "upa"
 FLD_KB_OBJ_NAME = "name"

src/loaders/common/loader_helper.py

@@ -1,4 +1,5 @@
 import argparse
+import configparser
 import itertools
 import json
 import os
@@ -28,6 +29,7 @@
 )
 from src.loaders.common.loader_common_names import (
     COLLECTION_SOURCE_DIR,
+    CONTAINERS_CONF_PARAMS,
     DOCKER_HOST,
     FATAL_ERROR,
     FATAL_STACKTRACE,
@@ -289,6 +291,37 @@
     return proc
 
 
+def _get_containers_config(conf_path: str):
+    """Get containers.conf file at home directory."""
+    config = configparser.ConfigParser()
+    config.read(conf_path)
+    return config
+
+
+def is_config_modification_required(conf_path: str):
+    """check if the config requires modification."""
+    config = _get_containers_config(conf_path)
+    if not config.has_section("containers"):
+        return True
+    for key, val in CONTAINERS_CONF_PARAMS.items():
+        if config.get("containers", key, fallback=None) != val:
+            return True
+    return False
+
+
+def setup_callback_server_logs(conf_path: str):
+    """Set up containers.conf file for the callback server logs."""
+    config = _get_containers_config(conf_path)
+
+    if not config.has_section("containers"):
+        config.add_section("containers")
+
+    for key, val in CONTAINERS_CONF_PARAMS.items():
+        config.set("containers", key, val)
+
+    return config
+
+
 def is_upa_info_complete(upa_dir: str):
     """
     Check whether an UPA needs to be downloaded or not by loading the metadata file.
@@ -421,19 +454,30 @@
     os.link(target_file, new_file)
 
 
-def list_objects(wsid, conf, object_type, include_metadata=False, batch_size=10000):
+def list_objects(wsid, ws, object_type, include_metadata=False, batch_size=10000):
     """
     List all objects information given a workspace ID.
+
+    Args:
+        wsid (int): Target workspace addressed by the permanent ID
+        ws (Workspace): Workspace client
+        object_type (str): Type of the objects to be listed
+        include_metadata (boolean): Whether to include the user provided metadata in the returned object_info
+        batch_size (int): Number of objects to process in each batch
+
+    Returns:
+        list: a list of objects on the target workspace
+
     """
     if batch_size > 10000:
         raise ValueError("Maximum value for listing workspace objects is 10000")
 
-    maxObjectID = conf.ws.get_workspace_info({"id": wsid})[4]
+    maxObjectID = ws.get_workspace_info({"id": wsid})[4]
     batch_input = [
         [idx + 1, idx + batch_size] for idx in range(0, maxObjectID, batch_size)
     ]
     objs = [
-        conf.ws.list_objects(
+        ws.list_objects(
             _list_objects_params(wsid, min_id, max_id, object_type, include_metadata)
         )
         for min_id, max_id in batch_input