[Containerapp] Support debug console

src/containerapp/HISTORY.rst

@@ -4,6 +4,7 @@ Release History
 ===============
 upcoming
 ++++++
+* 'az containerapp debug': Open an SSH-like interactive shell within a container app debug console.
 
 1.0.0b2
 ++++++

src/containerapp/azext_containerapp/_help.py

@@ -2054,3 +2054,12 @@
           az containerapp job registry set -n my-containerapp-job -g MyResourceGroup \\
               --server MyContainerappJobRegistry.azurecr.io --identity system-environment
 """
+
+helps['containerapp debug'] = """
+    type: command
+    short-summary: Open an SSH-like interactive shell within a container app debug console.
+    examples:
+    - name: debug by connecting to a container app's debug console by replica, revision and container
+      text: |
+          az containerapp debug -n MyContainerapp -g MyResourceGroup --revision MyRevision --replica MyReplica --container MyContainer
+"""

src/containerapp/azext_containerapp/_params.py

@@ -440,3 +440,13 @@ def load_arguments(self, _):
         c.argument('logger_name', help="The logger name.")
         c.argument('logger_level', arg_type=get_enum_type(["off", "error", "info", "debug", "trace", "warn"]), help="Set the log level for the specific logger name.")
         c.argument('all', help="The flag to indicate all logger settings.", action="store_true")
+
+    with self.argument_context('containerapp debug') as c:
+        c.argument('container',
+                   help="The container name that the debug console will connect to. Defaults to the first container of first replica.")
+        c.argument('replica',
+                   help="The name of the replica. List replicas with 'az containerapp replica list'. A replica may be not found when it's scaled to zero if there is no traffic to your app. Defaults to the first replica of 'az containerapp replica list'.")
+        c.argument('revision',
+                   help="The name of the container app revision. Defaults to the latest revision.")
+        c.argument('name', name_type, id_part=None, help="The name of the Containerapp.")
+        c.argument('resource_group_name', arg_type=resource_group_name_type, id_part=None)

src/containerapp/azext_containerapp/_ssh_utils.py

@@ -0,0 +1,43 @@
+# --------------------------------------------------------------------------------------------
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License. See License.txt in the project root for license information.
+# --------------------------------------------------------------------------------------------
+# pylint: disable=logging-fstring-interpolation
+# pylint: disable=possibly-used-before-assignment
+
+from azure.cli.command_modules.containerapp._ssh_utils import WebSocketConnection, SSH_TERM_RESIZE_PREFIX, \
+    SSH_DEFAULT_ENCODING, read_ssh
+from azure.cli.core.commands.client_factory import get_subscription_id
+
+from knack.log import get_logger
+
+logger = get_logger(__name__)
+
+
+class DebugWebSocketConnection(WebSocketConnection):
+    def __init__(self, cmd, resource_group_name, name, revision, replica, container):
+        super(DebugWebSocketConnection, self).__init__(cmd, resource_group_name, name, revision, replica, container, "")
+
+    def _get_url(self, cmd, resource_group_name, name, revision, replica, container, startup_command):
+        sub = get_subscription_id(cmd.cli_ctx)
+        base_url = self._logstream_endpoint
+        proxy_api_url = base_url[:base_url.index("/subscriptions/")].replace("https://", "")
+
+        return (f"wss://{proxy_api_url}/subscriptions/{sub}/resourceGroups/{resource_group_name}/containerApps/{name}"
+                f"/revisions/{revision}/replicas/{replica}/debug"
+                f"?targetContainer={container}")
+
+
+def read_debug_ssh(connection: WebSocketConnection, response_encodings):
+    from shutil import get_terminal_size
+    size = get_terminal_size()
+    if connection.is_connected:
+        # We need to send resize for the whole session two times.
+        # First time is here and second time is in `read_ssh` method.
+        # refer `kubectl debug` command implementation:
+        # https://github.com/kubernetes/kubectl/blob/14f6a11dd84315dc5179ff04156b338def935eaa/pkg/cmd/attach/attach.go#L296
+        connection.send(b"".join([SSH_TERM_RESIZE_PREFIX,
+                                  f'{{"Width": {size.columns + 1}, '
+                                  f'"Height": {size.lines}}}'.encode(SSH_DEFAULT_ENCODING)]))
+
+    read_ssh(connection, response_encodings)

src/containerapp/azext_containerapp/_validators.py

@@ -7,11 +7,16 @@
 from knack.log import get_logger
 from urllib.parse import urlparse
 
-from azure.cli.core.azclierror import (ValidationError, InvalidArgumentValueError,
-                                       MutuallyExclusiveArgumentError, RequiredArgumentMissingError)
-from azure.cli.command_modules.containerapp._utils import is_registry_msi_system
 from azure.mgmt.core.tools import is_valid_resource_id
 
+from azure.cli.command_modules.containerapp._utils import is_registry_msi_system, safe_get
+from azure.cli.command_modules.containerapp._validators import _validate_revision_exists, _validate_replica_exists, \
+    _validate_container_exists
+from azure.cli.core.azclierror import (InvalidArgumentValueError,
+                                       MutuallyExclusiveArgumentError, RequiredArgumentMissingError,
+                                       ResourceNotFoundError, ValidationError)
+
+from ._clients import ContainerAppPreviewClient
 from ._utils import is_registry_msi_system_environment
 
 from ._constants import ACR_IMAGE_SUFFIX, \
@@ -215,3 +220,49 @@ def validate_timeout_in_seconds(cmd, namespace):
     if timeout_in_seconds is not None:
         if timeout_in_seconds < 0 or timeout_in_seconds > 60:
             raise ValidationError("timeout in seconds must be in range [0, 60].")
+
+
+def validate_debug(cmd, namespace):
+    print("Validating...")
+    revision_already_set = bool(namespace.revision)
+    replica_already_set = bool(namespace.replica)
+    container_already_set = bool(namespace.container)
+    _set_debug_defaults(cmd, namespace)
+    if revision_already_set:
+        _validate_revision_exists(cmd, namespace)
+    if replica_already_set:
+        _validate_replica_exists(cmd, namespace)
+    if container_already_set:
+        _validate_container_exists(cmd, namespace)
+
+
+def _set_debug_defaults(cmd, namespace):
+    app = ContainerAppPreviewClient.show(cmd, namespace.resource_group_name, namespace.name)
+    if not app:
+        raise ResourceNotFoundError("Could not find a container app")
+    if not namespace.revision:
+        namespace.revision = app.get("properties", {}).get("latestRevisionName")
+        if not namespace.revision:
+            raise ResourceNotFoundError("Could not find a revision")
+    if not namespace.replica:
+        replicas = ContainerAppPreviewClient.list_replicas(
+            cmd=cmd,
+            resource_group_name=namespace.resource_group_name,
+            container_app_name=namespace.name,
+            revision_name=namespace.revision
+        )
+        if not replicas:
+            raise ResourceNotFoundError("Could not find a active replica")
+        namespace.replica = replicas[0]["name"]
+        if not namespace.container and replicas[0]["properties"]["containers"]:
+            namespace.container = replicas[0]["properties"]["containers"][0]["name"]
+    if not namespace.container:
+        revision = ContainerAppPreviewClient.show_revision(
+            cmd,
+            resource_group_name=namespace.resource_group_name,
+            container_app_name=namespace.name,
+            name=namespace.revision
+        )
+        revision_containers = safe_get(revision, "properties", "template", "containers")
+        if revision_containers:
+            namespace.container = revision_containers[0]["name"]

src/containerapp/azext_containerapp/commands.py

@@ -7,13 +7,15 @@
 # from azure.cli.core.commands import CliCommandType
 # from azure.mgmt.core.tools import is_valid_resource_id, parse_resource_id
 from azure.cli.command_modules.containerapp._transformers import (transform_containerapp_output, transform_containerapp_list_output)
+
 from azext_containerapp._client_factory import ex_handler_factory
 from ._transformers import (transform_sensitive_values,
                             transform_telemetry_data_dog_values,
                             transform_telemetry_app_insights_values,
                             transform_telemetry_otlp_values,
                             transform_telemetry_otlp_values_by_name_wrapper)
 from ._utils import is_cloud_supported_by_connected_env
+from ._validators import validate_debug
 
 
 def load_command_table(self, args):
@@ -24,6 +26,7 @@ def load_command_table(self, args):
         g.custom_command('update', 'update_containerapp', supports_no_wait=True, exception_handler=ex_handler_factory(), table_transformer=transform_containerapp_output, transform=transform_sensitive_values)
         g.custom_command('delete', 'delete_containerapp', supports_no_wait=True, confirmation=True, exception_handler=ex_handler_factory())
         g.custom_command('up', 'containerapp_up', supports_no_wait=False, exception_handler=ex_handler_factory())
+        g.custom_command('debug', 'containerapp_debug', is_preview=True, validator=validate_debug)
 
     with self.command_group('containerapp replica') as g:
         g.custom_show_command('show', 'get_replica')  # TODO implement the table transformer

src/containerapp/azext_containerapp/custom.py

@@ -4,13 +4,15 @@
 # --------------------------------------------------------------------------------------------
 # pylint: disable=line-too-long, unused-argument, logging-fstring-interpolation, logging-not-lazy, consider-using-f-string, logging-format-interpolation, inconsistent-return-statements, broad-except, bare-except, too-many-statements, too-many-locals, too-many-boolean-expressions, too-many-branches, too-many-nested-blocks, pointless-statement, expression-not-assigned, unbalanced-tuple-unpacking, unsupported-assignment-operation
 
+import threading
 import time
 from urllib.parse import urlparse
 import json
 import requests
 import subprocess
 from concurrent.futures import ThreadPoolExecutor
 
+from azure.cli.command_modules.containerapp._ssh_utils import SSH_BACKUP_ENCODING, SSH_CTRL_C_MSG, get_stdin_writer
 from azure.cli.core import telemetry as telemetry_core
 
 from azure.cli.core.azclierror import (
@@ -112,6 +114,8 @@
     AzureFileProperties as AzureFilePropertiesModel
 )
 
+from ._ssh_utils import (SSH_DEFAULT_ENCODING, DebugWebSocketConnection, read_debug_ssh)
+
 from ._utils import connected_env_check_cert_name_availability, get_oryx_run_image_tags, patchable_check, get_pack_exec_path, is_docker_running, parse_build_env_vars, env_has_managed_identity
 
 from ._constants import (CONTAINER_APPS_RP,
@@ -3224,3 +3228,36 @@ def set_registry_job(cmd, name, resource_group_name, server, username=None, pass
     containerapp_job_registry_set_decorator.construct_payload()
     r = containerapp_job_registry_set_decorator.set()
     return r
+
+
+def containerapp_debug(cmd, resource_group_name, name, container=None, revision=None, replica=None):
+    print("Connecting...")
+    conn = DebugWebSocketConnection(
+        cmd=cmd,
+        resource_group_name=resource_group_name,
+        name=name,
+        revision=revision,
+        replica=replica,
+        container=container
+    )
+
+    encodings = [SSH_DEFAULT_ENCODING, SSH_BACKUP_ENCODING]
+    reader = threading.Thread(target=read_debug_ssh, args=(conn, encodings))
+    reader.daemon = True
+    reader.start()
+
+    writer = get_stdin_writer(conn)
+    writer.daemon = True
+    writer.start()
+
+    while conn.is_connected:
+        if not reader.is_alive() or not writer.is_alive():
+            logger.warning("Reader or Writer for WebSocket is not alive. Closing the connection.")
+            conn.disconnect()
+
+        try:
+            time.sleep(0.1)
+        except KeyboardInterrupt:
+            if conn.is_connected:
+                logger.info("Caught KeyboardInterrupt. Sending ctrl+c to server")
+                conn.send(SSH_CTRL_C_MSG)