Add --container-name for custom docker container name

docs/cli.md

@@ -90,7 +90,8 @@ sb benchmark list-parameters --name pytorch-[a-z]+
 
 Deploy the SuperBench environments to all managed nodes.
 ```bash title="SB CLI"
-sb deploy [--docker-image]
+sb deploy [--container-name]
+          [--docker-image]
           [--docker-password]
           [--docker-username]
           [--host-file]
@@ -106,6 +107,7 @@ sb deploy [--docker-image]
 
 | Name                  | Default                 | Description                                                                       |
 |-----------------------|-------------------------|-----------------------------------------------------------------------------------|
+| `--container-name`    | `sb-workspace`          | Docker container name.                                                            |
 | `--docker-image` `-i` | `superbench/superbench` | Docker image URI, [here](./user-tutorial/container-images.mdx) listed all images. |
 | `--docker-password`   | `None`                  | Docker registry password if authentication is needed.                             |
 | `--docker-username`   | `None`                  | Docker registry username if authentication is needed.                             |
@@ -130,6 +132,11 @@ Deploy default image on local GPU node:
 sb deploy --host-list localhost
 ```
 
+Deploy using custom container name:
+```bash title="SB CLI"
+sb deploy --host-list localhost --container-name my-benchmark
+```
+
 Deploy image `superbench/cuda:11.1` to all nodes in `./host.ini`:
 ```bash title="SB CLI"
 sb deploy --docker-image superbench/cuda:11.1 --host-file ./host.ini
@@ -348,6 +355,7 @@ Run the SuperBench benchmarks distributedly.
 ```bash title="SB CLI"
 sb run [--config-file]
        [--config-override]
+       [--container-name]
        [--docker-image]
        [--docker-password]
        [--docker-username]
@@ -367,6 +375,7 @@ sb run [--config-file]
 |--------------------------|-------------------------|-----------------------------------------------------------------------------|
 | `--config-file` `-c`     | `None`                  | Path to SuperBench config file.                                             |
 | `--config-override` `-C` | `None`                  | Extra arguments to override config_file.                                    |
+| `--container-name`       | `sb-workspace`          | Docker container name.                                                      |
 | `--docker-image` `-i`    | `superbench/superbench` | Docker image URI.                                                           |
 | `--docker-password`      | `None`                  | Docker registry password if authentication is needed.                       |
 | `--docker-username`      | `None`                  | Docker registry username if authentication is needed.                       |
@@ -392,6 +401,11 @@ Run all benchmarks on local GPU node:
 sb run --host-list localhost
 ```
 
+Run all benchmarks using custom container name:
+```bash title="SB CLI"
+sb run --host-list localhost --container-name my-benchmark
+```
+
 Run all benchmarks on all managed nodes in `./host.ini` using image `superbench/cuda:11.1`
 and default benchmarking configuration:
 ```bash title="SB CLI"

superbench/cli/_commands.py

@@ -42,6 +42,12 @@ def load_arguments(self, command):
         """
         with ArgumentsContext(self, '') as ac:
             ac.argument('docker_image', options_list=('--docker-image', '-i'), type=str, help='Docker image URI.')
+            ac.argument(
+                'container_name',
+                options_list=('--container-name',),
+                type=str,
+                help='Docker container name. Defaults to sb-workspace.'
+            )
             ac.argument('docker_username', type=str, help='Docker registry username if authentication is needed.')
             ac.argument('docker_password', type=str, help='Docker registry password if authentication is needed.')
             ac.argument('no_docker', action='store_true', help='Run on host directly without Docker.')

superbench/cli/_handler.py

@@ -96,6 +96,7 @@ def process_config_arguments(config_file=None, config_override=None, output_dir=
 
 def process_runner_arguments(
     docker_image='superbench/superbench',
+    container_name='sb-workspace',
     docker_username=None,
     docker_password=None,
     no_docker=False,
@@ -113,6 +114,7 @@ def process_runner_arguments(
 
     Args:
         docker_image (str, optional): Docker image URI. Defaults to superbench/superbench:latest.
+        container_name (str, optional): Docker container name. Defaults to sb-workspace.
         docker_username (str, optional): Docker registry username if authentication is needed. Defaults to None.
         docker_password (str, optional): Docker registry password if authentication is needed. Defaults to None.
         no_docker (bool, optional): Run on host directly without Docker. Defaults to False.
@@ -138,6 +140,8 @@ def process_runner_arguments(
     """
     if bool(docker_username) != bool(docker_password):
         raise CLIError('Must specify both docker_username and docker_password if authentication is needed.')
+    if not container_name or not container_name.strip():
+        raise CLIError('container_name cannot be empty.')
     if not (host_file or host_list):
         raise CLIError('Must specify one of host_file or host_list.')
     host_file = check_argument_file('host_file', host_file)
@@ -147,6 +151,7 @@ def process_runner_arguments(
     docker_config = OmegaConf.create(
         {
             'image': docker_image,
+            'container_name': container_name.strip(),
             'username': docker_username,
             'password': docker_password,
             'registry': split_docker_domain(docker_image)[0],
@@ -210,6 +215,7 @@ def exec_command_handler(config_file=None, config_override=None, output_dir=None
 
 def deploy_command_handler(
     docker_image='superbench/superbench',
+    container_name='sb-workspace',
     docker_username=None,
     docker_password=None,
     no_image_pull=False,
@@ -230,6 +236,7 @@ def deploy_command_handler(
 
     Args:
         docker_image (str, optional): Docker image URI. Defaults to superbench/superbench:latest.
+        container_name (str, optional): Docker container name. Defaults to sb-workspace.
         docker_username (str, optional): Docker registry username if authentication is needed. Defaults to None.
         docker_password (str, optional): Docker registry password if authentication is needed. Defaults to None.
         no_image_pull (bool, optional): Skip pull and use local Docker image. Defaults to False.
@@ -245,6 +252,7 @@ def deploy_command_handler(
     """
     docker_config, ansible_config, sb_config, sb_output_dir = process_runner_arguments(
         docker_image=docker_image,
+        container_name=container_name,
         docker_username=docker_username,
         docker_password=docker_password,
         no_docker=False,
@@ -265,6 +273,7 @@ def deploy_command_handler(
 
 def run_command_handler(
     docker_image='superbench/superbench',
+    container_name='sb-workspace',
     docker_username=None,
     docker_password=None,
     no_docker=False,
@@ -284,6 +293,7 @@ def run_command_handler(
 
     Args:
         docker_image (str, optional): Docker image URI. Defaults to superbench/superbench:latest.
+        container_name (str, optional): Docker container name. Defaults to sb-workspace.
         docker_username (str, optional): Docker registry username if authentication is needed. Defaults to None.
         docker_password (str, optional): Docker registry password if authentication is needed. Defaults to None.
         no_docker (bool, optional): Run on host directly without Docker. Defaults to False.
@@ -303,6 +313,7 @@ def run_command_handler(
     """
     docker_config, ansible_config, sb_config, sb_output_dir = process_runner_arguments(
         docker_image=docker_image,
+        container_name=container_name,
         docker_username=docker_username,
         docker_password=docker_password,
         no_docker=no_docker,

superbench/cli/_help.py

@@ -34,6 +34,8 @@
     examples:
         - name: deploy default image on local GPU node
           text: {cli_name} deploy --host-list localhost
+        - name: deploy using custom container name
+          text: {cli_name} deploy --host-list localhost --container-name my-benchmark
         - name: deploy image "superbench/cuda:11.1" to all nodes in ./host.ini
           text: {cli_name} deploy --docker-image superbench/cuda:11.1 --host-file ./host.ini
         - name: deploy image "superbench/rocm:4.0" to node-0 and node-2, using key file id_rsa for ssh
@@ -56,6 +58,8 @@
     examples:
         - name: run all benchmarks on local GPU node
           text: {cli_name} run --host-list localhost
+        - name: run all benchmarks using custom container name
+          text: {cli_name} run --host-list localhost --container-name my-benchmark
         - name: run all benchmarks on all nodes in ./host.ini using image "superbench/cuda:11.1"
             and default benchmarking configuration
           text: {cli_name} run --docker-image superbench/cuda:11.1 --host-file ./host.ini

superbench/runner/runner.py

@@ -52,6 +52,11 @@ def __init__(self, sb_config, docker_config, ansible_config, sb_output_dir):
         self._sb_enabled_benchmarks = self.__get_enabled_benchmarks()
         logger.info('Runner will run: %s', self._sb_enabled_benchmarks)
 
+    @property
+    def _container_name(self):
+        """Get docker container name with backward-compatible default."""
+        return getattr(self._docker_config, 'container_name', 'sb-workspace')
+
     def __set_logger(self, filename):
         """Set logger and add file handler.
 
@@ -268,6 +273,7 @@ def deploy(self):    # pragma: no cover
         extravars = {
             'ssh_port': random.randint(1 << 14, (1 << 15) - 1),
             'output_dir': str(self._output_path),
+            'container': self._container_name,
             'docker_image': self._docker_config.image,
             'docker_pull': bool(self._docker_config.pull),
         }
@@ -287,7 +293,7 @@ def run_sys_info(self):
 
         logger.info('Runner is going to get node system info.')
 
-        fcmd = "docker exec sb-workspace bash -lc '{command}'"
+        fcmd = "docker exec {container} bash -lc '{{command}}'".format(container=self._container_name)
 
         if 'skip' not in self._docker_config:
             self._docker_config.skip = False
@@ -310,6 +316,7 @@ def check_env(self):    # pragma: no cover
             self._ansible_client.get_playbook_config(
                 'check_env.yaml',
                 extravars={
+                    'container': self._container_name,
                     'no_docker': False if 'skip' not in self._docker_config else self._docker_config.skip,
                     'output_dir': str(self._output_path),
                     'env': '\n'.join(f'{k}={v}' for k, v in self._sb_config.superbench.env.items()),
@@ -550,7 +557,7 @@ def _run_proc(self, benchmark_name, mode, vars):
         if mode_env_cmds:
             env_list = f"{env_list} && {' && '.join(mode_env_cmds)}"
 
-        fcmd = "docker exec sb-workspace bash -lc '{env_list} && {command}'"
+        fcmd = "docker exec {container} bash -lc '{{env_list}} && {{command}}'".format(container=self._container_name)
         if self._docker_config.skip:
             fcmd = "bash -c '{env_list} && cd $SB_WORKSPACE && {command}'"
         ansible_runner_config = self._ansible_client.get_shell_config(

tests/cli/test_handler.py

@@ -5,6 +5,8 @@
 
 import unittest
 
+from knack.util import CLIError
+
 import superbench.cli._handler as cli_handler
 
 
@@ -78,3 +80,19 @@ def test_split_docker_domain(self):
                 domain, name = cli_handler.split_docker_domain(test_case['input'])
                 self.assertEqual(domain, test_case['domain'])
                 self.assertEqual(name, test_case['name'])
+
+    def test_process_runner_arguments_container_name(self):
+        """Test process_runner_arguments saves custom container name."""
+        docker_config, _, _, _ = cli_handler.process_runner_arguments(
+            host_list='localhost',
+            container_name='sb-custom',
+        )
+        self.assertEqual(docker_config.container_name, 'sb-custom')
+
+    def test_process_runner_arguments_empty_container_name(self):
+        """Test process_runner_arguments rejects empty container name."""
+        with self.assertRaises(CLIError):
+            cli_handler.process_runner_arguments(
+                host_list='localhost',
+                container_name='   ',
+            )

tests/runner/test_runner.py

@@ -511,6 +511,30 @@ def fake_get_shell_config(cmd):
         self.assertIn('-x PATH', captured['cmd'])
         self.assertNotIn('export "PATH=', captured['cmd'])
 
+    @mock.patch('superbench.runner.ansible.AnsibleClient.run')
+    def test_run_proc_uses_custom_container_name(self, mock_ansible_client_run):
+        """Test _run_proc uses configured docker container name."""
+        mock_ansible_client_run.return_value = 0
+        self.runner._sb_benchmarks = {'foo': {}}
+        self.runner._docker_config.container_name = 'sb-custom'
+        captured = {}
+
+        def fake_get_shell_config(cmd):
+            captured['cmd'] = cmd
+            return {'module_args': cmd, 'cmdline': '', 'host_pattern': 'localhost', 'module': 'shell'}
+
+        self.runner._ansible_client.get_shell_config = fake_get_shell_config
+        mode = OmegaConf.create({
+            'name': 'local',
+            'proc_num': 1,
+            'env': {},
+            'prefix': '',
+        })
+
+        self.runner._run_proc('foo', mode, {'proc_rank': 0})
+
+        self.assertIn('docker exec sb-custom bash -lc', captured['cmd'])
+
     @mock.patch('superbench.runner.ansible.AnsibleClient.run')
     def test_run_proc_no_docker_keeps_tmp_env_source(self, mock_ansible_client_run):
         """Test _run_proc still sources /tmp/sb.env in no_docker mode."""