From c202718b590cdb80521505206a27a0ffcc7d4d6c Mon Sep 17 00:00:00 2001 From: YAN Songyang Date: Tue, 3 Feb 2026 16:25:42 +0900 Subject: [PATCH 1/5] Add gitignore --- .gitignore | 52 ++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) create mode 100644 .gitignore diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..2e7e106 --- /dev/null +++ b/.gitignore @@ -0,0 +1,52 @@ +# Python +__pycache__/ +*.py[cod] +*$py.class +.pytest_cache/ +.mypy_cache/ +.ruff_cache/ +.coverage +coverage.xml +htmlcov/ + +# Virtualenv / Conda +.venv/ +venv/ +ENV/ +env/ +.env +.conda/ +conda-bld/ + +# Jupyter +.ipynb_checkpoints/ + +# OS / editor +.DS_Store +Thumbs.db +.idea/ +.vscode/ +.cursor/ + +# Logs +*.log + +# Build / dist +build/ +dist/ +*.egg-info/ + +# ApolloSimFuzz outputs +results/ + +# Local clone of Apollo +apollo/ + +# TrafficSandbox generated artifacts / caches +TrafficSandbox/**/__pycache__/ +TrafficSandbox/**/.pytest_cache/ +TrafficSandbox/logs/ +TrafficSandbox/map_toolkit/data/**/map.pickle + +# Large map assets that are user/environment specific +TrafficSandbox/map_toolkit/data/**/map.json From f104d724c9c0e3161b6ef58457e79a43bf6e3201 Mon Sep 17 00:00:00 2001 From: YAN Songyang Date: Tue, 3 Feb 2026 16:28:11 +0900 Subject: [PATCH 2/5] Cap CPU allocation in create_apollo_ctn_configs based on host CPU count --- scenario_runner/ctn_manager.py | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/scenario_runner/ctn_manager.py b/scenario_runner/ctn_manager.py index e96f41e..5132bef 100644 --- a/scenario_runner/ctn_manager.py +++ b/scenario_runner/ctn_manager.py @@ -1,3 +1,4 @@ +import os from dataclasses import dataclass from loguru import logger from typing import List @@ -35,6 +36,11 @@ def create_apollo_ctn_configs( num_gpus = len(available_gpus) logger.info(f"Setting up {apollo_ctn_num} apollo containers for parallel execution with AVAILABLE {num_gpus} GPUs: {available_gpus}") + # Docker validates --cpus against the host CPU count. Cap requested CPU so + # we don't fail on smaller machines (e.g. 12-core host cannot run --cpus 24.0). + host_cpus = float(os.cpu_count() or 1) + cpu_cap = min(24.0, host_cpus) + operators = [] for i in range(apollo_ctn_num): # If there is only one GPU, assign all containers to it @@ -48,7 +54,8 @@ def create_apollo_ctn_configs( idx=i, container_name=f"apollo_{run_tag}_{i}", gpu=gpu_id, - cpu='24.0', # TODO: move to config + # TODO: move to config + cpu=str(cpu_cap), # capped by host CPU count apollo_root=apollo_root, dreamview_port=dreamview_port, bridge_port=bridge_port, From c9f174c488971cadd952329b16b4c80159fa5228 Mon Sep 17 00:00:00 2001 From: YAN Songyang Date: Tue, 3 Feb 2026 16:28:42 +0900 Subject: [PATCH 3/5] Enhance IP address retrieval in SandboxContainer and SandboxOperator to support newer Docker versions and fallback for host-networked containers. --- scenario_runner/sandbox_operator.py | 24 ++++++++++++++++++++++-- 1 file changed, 22 insertions(+), 2 deletions(-) diff --git a/scenario_runner/sandbox_operator.py b/scenario_runner/sandbox_operator.py index 4b9304d..f683819 100644 --- a/scenario_runner/sandbox_operator.py +++ b/scenario_runner/sandbox_operator.py @@ -44,7 +44,18 @@ def host(self): time.sleep(1.0) ctn = docker.from_env().containers.get(self.container_name) - return ctn.attrs['NetworkSettings']['IPAddress'] + ns = ctn.attrs.get("NetworkSettings") or {} + # Newer Docker versions may not populate top-level "IPAddress"; prefer Networks. + ip = ns.get("IPAddress") + if ip: + return ip + networks = ns.get("Networks") or {} + for net in networks.values(): + ip = net.get("IPAddress") + if ip: + return ip + # Fallback for host-networked containers (or unexpected daemon payloads) + return "localhost" @property def is_running(self) -> bool: @@ -152,7 +163,16 @@ def host(self) -> str: if not self.is_running: raise RuntimeError(f"Container {self.container_name} is not running.") ctn = docker.from_env().containers.get(self.container_name) - return ctn.attrs["NetworkSettings"]["IPAddress"] + ns = ctn.attrs.get("NetworkSettings") or {} + ip = ns.get("IPAddress") + if ip: + return ip + networks = ns.get("Networks") or {} + for net in networks.values(): + ip = net.get("IPAddress") + if ip: + return ip + return "localhost" # ===================== # Connection handling From 8dee45e61d747694687d354dff505fe71749ca83 Mon Sep 17 00:00:00 2001 From: YAN Songyang Date: Tue, 3 Feb 2026 16:29:53 +0900 Subject: [PATCH 4/5] Update configuration and enhance ApolloContainer for improved Docker integration. Especially the container name issue --- apollo_bridge/common/container.py | 108 +++++++++++++++++------------- 1 file changed, 63 insertions(+), 45 deletions(-) diff --git a/apollo_bridge/common/container.py b/apollo_bridge/common/container.py index 1a48707..85b8096 100644 --- a/apollo_bridge/common/container.py +++ b/apollo_bridge/common/container.py @@ -33,9 +33,19 @@ def __init__( self.APOLLO_MODULES = modules + # NOTE: Apollo's dev start scripts create a docker container named: + # apollo_dev_${USER} + # In this project, we set USER= when starting containers, + # so the actual docker container is `apollo_dev_`. + # Keep `instance_name` for logging, but always use `container_name` + # for docker API calls / docker exec / docker start. self.user = name - self.name = name - # self.name = f"apollo_dev_{self.user}" # test for name + self.instance_name = name + # Backward-compat: other parts of the codebase expect `.name` to exist. + # Here `.name` refers to the logical instance name (e.g. "apollo_debug_0"), + # not the docker container name. + self.name = self.instance_name + self.container_name = f"apollo_dev_{self.user}" # create docker container if it is not exist self.hd_map = map_name self.apollo_root = apollo_root @@ -64,11 +74,17 @@ def host(self) -> str: :type: str """ - ctn = docker.from_env().containers.get(self.name) - host = ctn.attrs['NetworkSettings']['IPAddress'] - if host == '': - return 'localhost' - return ctn.attrs['NetworkSettings']['IPAddress'] + ctn = docker.from_env().containers.get(self.container_name) + ns = ctn.attrs.get("NetworkSettings") or {} + ip = ns.get("IPAddress") + if ip: + return ip + networks = ns.get("Networks") or {} + for net in networks.values(): + ip = net.get("IPAddress") + if ip: + return ip + return "localhost" @property def is_container_running(self) -> bool: @@ -79,7 +95,7 @@ def is_container_running(self) -> bool: :rtype: bool """ try: - return docker.from_env().containers.get(self.name).status == 'running' + return docker.from_env().containers.get(self.container_name).status == 'running' except Exception: return False @@ -118,14 +134,14 @@ def create_container(self): client = docker.from_env() has_create = False try: - container = client.containers.get(self.name) + _ = client.containers.get(self.container_name) has_create = True - # logger.debug(f'Apollo instance {self.name} already exists') + # logger.debug(f'Apollo instance {self.container_name} already exists') except Exception as e: pass if not has_create: - logger.info(f'Create Apollo container {self.name}') + logger.info(f'Create Apollo container {self.container_name}') # start_script_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'scripts', 'dev_start_ctn.sh') start_script_path = os.path.join(self.apollo_root, "docker", "scripts", "dev_start_ctn.sh") @@ -148,11 +164,11 @@ def create_container(self): def start_container(self): # while not self.is_container_running: if self.is_container_running: - logger.debug(f'Apollo container {self.name} is already running') + logger.debug(f'Apollo container {self.container_name} is already running') return True - logger.info(f'Start Apollo container {self.name}') - cmd = f'docker start {self.name}' + logger.info(f'Start Apollo container {self.container_name}') + cmd = f'docker start {self.container_name}' _ = subprocess.run(cmd, shell=True) time.sleep(0.5) self.start_time = time.time() @@ -161,8 +177,8 @@ def restart_container(self): """ Restarts an Apollo container instance """ - logger.info(f'Restart Apollo container {self.name}') - cmd = f'docker restart {self.name}' + logger.info(f'Restart Apollo container {self.container_name}') + cmd = f'docker restart {self.container_name}' _ = subprocess.run(cmd, shell=True) time.sleep(0.5) self.start_time = time.time() @@ -174,8 +190,8 @@ def stop_container(self): param bool restart: force container to restart """ if self.is_container_running: - logger.info(f'Stop Apollo container {self.name}') - cmd = f'docker stop {self.name}' + logger.info(f'Stop Apollo container {self.container_name}') + cmd = f'docker stop {self.container_name}' _ = subprocess.run(cmd, shell=True) ##### Apollo Cyber Bridge ##### @@ -187,7 +203,7 @@ def start_bridge(self) -> bool: for _ in range(10): # try 10 times max: try: - cmd = f"docker exec --user {self.user} -d {self.name} ./scripts/bridge.sh" + cmd = f"docker exec --user {self.user} -d {self.container_name} ./scripts/bridge.sh" _ = subprocess.run(cmd, shell=True) self.bridge = CyberBridge(self.host, self.bridge_port) break @@ -255,7 +271,9 @@ def start_dreamview( self, hd_map, dv_mode="Mkz Standard Debug", - apollo_type="Mkz_Example", + # Must match a folder under /apollo/modules/calibration/data/. + # In Apollo 7.0 this is `mkz_example` (lowercase) rather than `Mkz_Example`. + apollo_type="mkz_example", max_retry: int = 10, retry_interval: float = 1.0, ) -> bool: @@ -287,7 +305,7 @@ def try_connect() -> bool: # restart dreamview cmd = ( - f"docker exec --user {self.user} {self.name} " + f"docker exec --user {self.user} {self.container_name} " f"./scripts/bootstrap.sh restart --gpu {self.gpu_usage}" ) r = subprocess.run(cmd, shell=True, stdout=subprocess.DEVNULL, stderr=subprocess.PIPE) @@ -335,7 +353,7 @@ def start_modules_script(self): for launch in MODULE_LAUNCHES: cmd = ( - f'docker exec --user {self.user} -d {self.name} ' + f'docker exec --user {self.user} -d {self.container_name} ' f'bash -c "' f'source /apollo/scripts/apollo_base.sh && ' f'export CUDA_VISIBLE_DEVICES={self.gpu_usage} ' @@ -361,7 +379,7 @@ def disable_modules_script(self): for launch in MODULE_LAUNCHES: cmd = ( - f'docker exec --user {self.user} {self.name} ' + f'docker exec --user {self.user} {self.container_name} ' f'bash -c "' f'source /apollo/scripts/apollo_base.sh && ' f'cyber_launch stop {launch}' @@ -379,21 +397,21 @@ def start_recorder(self, record_folder: str, record_id: str): logger.info(f'Start Apollo recorder: {record_folder}/{record_id}') # cmd = f"docker exec --user {self.user} {self.name} rm -rf cyber_recorder.log.INFO*" - cmd = f"docker exec --user {self.user} {self.name} sh -c 'find /apollo -name \"cyber_recorder.log.INFO.*\" -delete'" + cmd = f"docker exec --user {self.user} {self.container_name} sh -c 'find /apollo -name \"cyber_recorder.log.INFO.*\" -delete'" # logger.debug(cmd) _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} rm -rf {record_folder}/{record_id}" + cmd = f"docker exec --user {self.user} {self.container_name} rm -rf {record_folder}/{record_id}" # logger.debug(cmd) _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} mkdir -p {record_folder}/{record_id}" + cmd = f"docker exec --user {self.user} {self.container_name} mkdir -p {record_folder}/{record_id}" # logger.debug(cmd) _ = subprocess.run(cmd, shell=True) container_cmd_recorder = "/apollo/bazel-bin/cyber/tools/cyber_recorder/cyber_recorder" container_cmd_cmd = f"{container_cmd_recorder} record -o {record_folder}/{record_id}/recording -a &" - cmd = f"docker exec -d --user {self.user} {self.name} {container_cmd_cmd}" + cmd = f"docker exec -d --user {self.user} {self.container_name} {container_cmd_cmd}" # logger.debug(cmd) _ = subprocess.run(cmd, shell=True) time.sleep(1.0) @@ -404,17 +422,17 @@ def stop_recorder(self): """ logger.info(f'Stop Apollo recorder.') container_cmd = "python3 /apollo/scripts/record_bag.py --stop --stop_signal SIGINT" - cmd = f"docker exec --user {self.user} {self.name} {container_cmd}" + cmd = f"docker exec --user {self.user} {self.container_name} {container_cmd}" # cmd = f"docker exec {self.name} {container_cmd}" _ = subprocess.run(cmd, shell=True) time.sleep(1.0) def copy_record(self, record_folder: str, record_id: str, target_folder: str, delete=False): - logger.info(f'Copy Apollo record: {self.name}:{record_folder}/{record_id} {target_folder}') - cmd = f'docker cp {self.name}:{record_folder}/{record_id} {target_folder}' + logger.info(f'Copy Apollo record: {self.container_name}:{record_folder}/{record_id} {target_folder}') + cmd = f'docker cp {self.container_name}:{record_folder}/{record_id} {target_folder}' _ = subprocess.run(cmd, shell=True) if delete: - cmd = f'docker exec --user {self.user} {self.name} rm -rf {record_folder}/{record_id}' + cmd = f'docker exec --user {self.user} {self.container_name} rm -rf {record_folder}/{record_id}' # cmd = f'docker exec {self.name} rm -rf {record_folder}/{record_id}' _ = subprocess.run(cmd, shell=True) @@ -423,22 +441,22 @@ def clean_cache(self): """ Removes Apollo's log files to save disk space """ - cmd = f"docker exec --user {self.user} {self.name} rm -rf /apollo/data" + cmd = f"docker exec --user {self.user} {self.container_name} rm -rf /apollo/data" _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} rm -rf /apollo/records" + cmd = f"docker exec --user {self.user} {self.container_name} rm -rf /apollo/records" _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} sh -c 'find /apollo -name \"*.log.*\" -delete'" + cmd = f"docker exec --user {self.user} {self.container_name} sh -c 'find /apollo -name \"*.log.*\" -delete'" _ = subprocess.run(cmd, shell=True) # create data dir - cmd = f"docker exec --user {self.user} {self.name} mkdir -p /apollo/data" + cmd = f"docker exec --user {self.user} {self.container_name} mkdir -p /apollo/data" _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} mkdir -p /apollo/data/bag" + cmd = f"docker exec --user {self.user} {self.container_name} mkdir -p /apollo/data/bag" _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} mkdir -p /apollo/data/log" + cmd = f"docker exec --user {self.user} {self.container_name} mkdir -p /apollo/data/log" _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} mkdir -p /apollo/data/core" + cmd = f"docker exec --user {self.user} {self.container_name} mkdir -p /apollo/data/core" _ = subprocess.run(cmd, shell=True) - cmd = f"docker exec --user {self.user} {self.name} mkdir -p /apollo/records" + cmd = f"docker exec --user {self.user} {self.container_name} mkdir -p /apollo/records" _ = subprocess.run(cmd, shell=True) ###### Global ###### @@ -469,7 +487,7 @@ def _start_modules_with_retry( return False def safe_shutdown(self): - logger.warning(f"Entering safe shutdown for container: {self.name}") + logger.warning(f"Entering safe shutdown for container: {self.container_name}") try: self.stop_recorder() except Exception: @@ -481,7 +499,7 @@ def safe_shutdown(self): pass def start(self) -> bool: - logger.info(f"Starting Apollo container: {self.name}") + logger.info(f"Starting Apollo container: {self.container_name}") # 1️⃣ 启动 container self.start_container() @@ -496,7 +514,7 @@ def start(self) -> bool: # 4️⃣ container 级别 retry（重操作） for i in range(5): logger.warning( - f"[{i+1}/5] Restart Apollo container due to Dreamview failure: {self.name}" + f"[{i+1}/5] Restart Apollo container due to Dreamview failure: {self.container_name}" ) self.restart_container() self.clean_cache() @@ -505,10 +523,10 @@ def start(self) -> bool: break else: # ❗ 最终失败，进入安全状态 - logger.error(f"Dreamview failed after container retries: {self.name}") + logger.error(f"Dreamview failed after container retries: {self.container_name}") self.safe_shutdown() raise RuntimeError( - f"Dreamview not running after retries: {self.name}" + f"Dreamview not running after retries: {self.container_name}" ) # 5️⃣ Dreamview OK，开始 modules @@ -520,7 +538,7 @@ def start(self) -> bool: self.safe_shutdown() raise RuntimeError("Apollo modules not running") - logger.info(f"Apollo container {self.name} started successfully") + logger.info(f"Apollo container {self.container_name} started successfully") return True From ae7d4789edfaacef8c347fbe78d734a1b6ed3e2e Mon Sep 17 00:00:00 2001 From: YAN Songyang Date: Tue, 3 Feb 2026 16:30:43 +0900 Subject: [PATCH 5/5] Add deap and ray to requirements.txt --- requirements.txt | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/requirements.txt b/requirements.txt index 654ab23..4f536ff 100644 --- a/requirements.txt +++ b/requirements.txt @@ -20,4 +20,6 @@ nvitop tabulate opencv-python matplotlib -natsort \ No newline at end of file +natsort +deap +ray \ No newline at end of file