Adaptation for API v5 revisions (Phase 1) (#97)

* feat: Add background task support
* feat: Implement progress bar for rescanning images using bgtask API
* refactor: Clean up BaseFunction inheritance using contextvars
  - Previously, to bind the current session with API function classes,
    we generated new type objects at runtime.
    - This has confused IDEs and type checkers.
    - Now type checkers can statically deduce the types for individual API
      function classes.
  - Now we use contextvars (ai.backend.client.session.api_session)
    to keep the reference to the current session.
    - There are *NO* public Session/AsyncSession API changes!
    - Only the API function classes need to be rewritten.
    - For synchronous Session, we pass the context to the separate worker
    thread using copy_context() whenever calling API functions, which is
    a light-weight operation.
* fix: Remove redundant src/ai/backend/client/etcd.py which had been already
  copied to src/ai/backend/client/func/etcd.py
* ci: Bump the lint Python version to 3.7
* ci: Remove Python 3.6 from AppVeyor
* ci: Make Python 3.8 as the officially supported environment by removing it
  from "allow_failures" environments in Travis CI
* ci: Now type errors are no longer ignored, and all remaining type errors are
  now fixed.
* test: All test cases in tests/cli/test_cli_proxy.py are marked "xfail"
  because there is an upstream issue rendering those tests always
  failing while real-world use cases have no problems.
  - ref) pytest-dev/pytest-asyncio#153

BREAKING-CHANGE: Dropped Python 3.6 support. Now it requires Python 3.7 or higher.

Author: achimnol

.github/workflows/lint-flake8.yml

@@ -9,10 +9,10 @@ jobs:
 
     steps:
     - uses: actions/checkout@v1
-    - name: Set up Python 3.6
+    - name: Set up Python 3.7
       uses: actions/setup-python@v1
       with:
-        python-version: 3.6
+        python-version: 3.7
     - name: Install dependencies
       run: |
         python -m pip install -U pip setuptools

.github/workflows/typecheck-mypy.yml

@@ -24,4 +24,4 @@ jobs:
         else
           echo "::add-matcher::.github/workflows/mypy-matcher.json"
         fi
-        python -m mypy --no-color-output src/ai/backend || exit 0
+        python -m mypy --no-color-output src/ai/backend

.travis.yml

@@ -10,7 +10,6 @@ stages:
 
 # build matrix for test stage
 python:
-  - "3.6"
   - "3.7"
   - "3.8"
 os:
@@ -47,13 +46,6 @@ jobs:
         username: "__token__"
         password:
           secure: "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"
-  allow_failures:
-    - python: "3.8"
-      os: linux
-  fast_finish: true
-  # exclude the duplicate default test stage
-  exclude:
-    - python: "3.6"
 
 notifications:
   webhooks:

appveyor.yml

@@ -1,8 +1,6 @@
 version: 1.0.dev{build}
 environment:
   matrix:
-    - PYTHON: "C:\\Python36"
-    - PYTHON: "C:\\Python36-x64"
     - PYTHON: "C:\\Python37"
     - PYTHON: "C:\\Python37-x64"
     - PYTHON: "C:\\Python38"

changes/97.breaking

@@ -0,0 +1 @@
+Drop support for Python 3.6

changes/97.feature

@@ -0,0 +1 @@
+Support APIv5's new GraphQL schema and kernel/session naming changes

setup.cfg

@@ -13,3 +13,7 @@ norecursedirs = venv virtualenv .git
 timeout = 5
 markers =
     integration: Test cases that require real manager (and agents) to be running on http://localhost:8081.
+
+[mypy]
+ignore_missing_imports = true
+namespace_packages = true

setup.py

@@ -20,6 +20,7 @@
     'tabulate~=0.8.6',
     'tqdm~=4.42',
     'yarl~=1.4.2',
+    'typing-extensions~=3.7.4',
 ]
 build_requires = [
     'wheel>=0.34.2',
@@ -82,7 +83,6 @@ def read_src_version():
         'Intended Audience :: Developers',
         'Programming Language :: Python',
         'Programming Language :: Python :: 3',
-        'Programming Language :: Python :: 3.6',
         'Programming Language :: Python :: 3.7',
         'Programming Language :: Python :: 3.8',
         'Operating System :: POSIX',
@@ -94,7 +94,7 @@ def read_src_version():
     ],
     package_dir={'': 'src'},
     packages=find_namespace_packages(where='src', include='ai.backend.*'),
-    python_requires='>=3.6',
+    python_requires='>=3.7',
     setup_requires=setup_requires,
     install_requires=install_requires,
     extras_require={

src/ai/backend/client/cli/admin/images.py

@@ -1,18 +1,22 @@
+import json
 import sys
+
 import click
 from tabulate import tabulate
+from tqdm import tqdm
 
 from . import admin
-from ...session import Session
+from ...compat import asyncio_run
+from ...session import Session, AsyncSession
 from ..pretty import print_done, print_warn, print_fail, print_error
 
 
 @admin.command()
 @click.option('--operation', is_flag=True, help='Get operational images only')
-def images(operation):
-    '''
+def images(operation: bool) -> None:
+    """
     Show the list of registered images in this cluster.
-    '''
+    """
     fields = [
         ('Name', 'name'),
         ('Registry', 'registry'),
@@ -32,33 +36,61 @@ def images(operation):
             print_warn('There are no registered images.')
             return
         print(tabulate((item.values() for item in items),
-                       headers=(item[0] for item in fields),
+                       headers=[item[0] for item in fields],
                        floatfmt=',.0f'))
 
 
 @admin.command()
 @click.option('-r', '--registry', type=str, default=None,
               help='The name (usually hostname or "lablup") '
                    'of the Docker registry configured.')
-def rescan_images(registry):
-    '''Update the kernel image metadata from all configured docker registries.'''
-    with Session() as session:
-        try:
-            result = session.Image.rescan_images(registry)
-        except Exception as e:
-            print_error(e)
-            sys.exit(1)
-        if result['ok']:
-            print_done("Updated the image metadata from the configured registries.")
-        else:
-            print_fail(f"Rescanning has failed: {result['msg']}")
+def rescan_images(registry: str) -> None:
+    """
+    Update the kernel image metadata from all configured docker registries.
+    """
+
+    async def rescan_images_impl(registry: str) -> None:
+        async with AsyncSession() as session:
+            try:
+                result = await session.Image.rescan_images(registry)
+            except Exception as e:
+                print_error(e)
+                sys.exit(1)
+            if not result['ok']:
+                print_fail(f"Failed to begin registry scanning: {result['msg']}")
+                sys.exit(1)
+            print_done("Started updating the image metadata from the configured registries.")
+            task_id = result['task_id']
+            bgtask = session.BackgroundTask(task_id)
+            try:
+                completion_msg_func = lambda: print_done("Finished registry scanning.")
+                with tqdm(unit='image') as pbar:
+                    async with bgtask.listen_events() as response:
+                        async for ev in response:
+                            data = json.loads(ev.data)
+                            if ev.event == 'task_updated':
+                                pbar.total = data['total_progress']
+                                pbar.write(data['message'])
+                                pbar.update(data['current_progress'] - pbar.n)
+                            elif ev.event == 'task_failed':
+                                error_msg = data['message']
+                                completion_msg_func = \
+                                    lambda: print_fail(f"Error occurred: {error_msg}")
+                            elif ev.event == 'task_cancelled':
+                                completion_msg_func = \
+                                    lambda: print_warn("Registry scanning has been "
+                                                       "cancelled in the middle.")
+            finally:
+                completion_msg_func()
+
+    asyncio_run(rescan_images_impl(registry))
 
 
 @admin.command()
 @click.argument('alias', type=str)
 @click.argument('target', type=str)
 def alias_image(alias, target):
-    '''Add an image alias.'''
+    """Add an image alias."""
     with Session() as session:
         try:
             result = session.Image.alias_image(alias, target)
@@ -74,7 +106,7 @@ def alias_image(alias, target):
 @admin.command()
 @click.argument('alias', type=str)
 def dealias_image(alias):
-    '''Remove an image alias.'''
+    """Remove an image alias."""
     with Session() as session:
         try:
             result = session.Image.dealias_image(alias)

src/ai/backend/client/cli/app.py

@@ -128,7 +128,7 @@ class ProxyRunnerContext:
     protocol: str
     host: str
     port: int
-    args: Dict[str, str]
+    args: Dict[str, Union[None, str, List[str]]]
     envs: Dict[str, str]
     api_session: Optional[AsyncSession]
     local_server: Optional[asyncio.AbstractServer]
@@ -150,7 +150,7 @@ def __init__(self, host: str, port: int,
         self.exit_code = 0
 
         self.args, self.envs = {}, {}
-        if len(args) > 0:
+        if args is not None and len(args) > 0:
             for argline in args:
                 tokens = []
                 for token in shlex.shlex(argline,
@@ -168,7 +168,7 @@ def __init__(self, host: str, port: int,
                     self.args[tokens[0]] = tokens[1]
                 else:
                     self.args[tokens[0]] = tokens[1:]
-        if len(envs) > 0:
+        if envs is not None and len(envs) > 0:
             for envline in envs:
                 split = envline.strip().split('=', maxsplit=2)
                 if len(split) == 2:
@@ -178,6 +178,7 @@ def __init__(self, host: str, port: int,
 
     async def handle_connection(self, reader: asyncio.StreamReader,
                                 writer: asyncio.StreamWriter) -> None:
+        assert self.api_session is not None
         p = WSProxy(self.api_session, self.session_name,
                     self.app_name, self.protocol,
                     self.args, self.envs,
@@ -232,6 +233,7 @@ async def __aexit__(self, *exc_info) -> None:
             print_info("Shutting down....")
             self.local_server.close()
             await self.local_server.wait_closed()
+        assert self.api_session is not None
         await self.api_session.__aexit__(*exc_info)
         assert self.api_session.closed
         if self.local_server is not None: