def main() -> NoReturn: clear_apport_folder() # Clear apport folder on start, otherwise duplicate crashes won't register initial_tombstones = set(get_tombstones()) sentry_sdk.utils.MAX_STRING_LENGTH = 8192 sentry_sdk.init("https://[email protected]/157615", default_integrations=False, release=get_version()) dongle_id = Params().get("DongleId", encoding='utf-8') sentry_sdk.set_user({"id": dongle_id}) sentry_sdk.set_tag("dirty", is_dirty()) sentry_sdk.set_tag("origin", get_origin()) sentry_sdk.set_tag("branch", get_branch()) sentry_sdk.set_tag("commit", get_commit()) sentry_sdk.set_tag("device", HARDWARE.get_device_type()) while True: now_tombstones = set(get_tombstones()) for fn, _ in (now_tombstones - initial_tombstones): try: cloudlog.info(f"reporting new tombstone {fn}") if fn.endswith(".crash"): report_tombstone_apport(fn) else: report_tombstone_android(fn) except Exception: cloudlog.exception(f"Error reporting tombstone {fn}") initial_tombstones = now_tombstones time.sleep(5)
def init(project: SentryProject) -> None: # forks like to mess with this, so double check comma_remote = is_comma_remote() and "commaai" in get_origin(default="") if not comma_remote or not is_registered_device() or PC: return env = "release" if is_tested_branch() else "master" dongle_id = Params().get("DongleId", encoding='utf-8') integrations = [] if project == SentryProject.SELFDRIVE: integrations.append(ThreadingIntegration(propagate_hub=True)) else: sentry_sdk.utils.MAX_STRING_LENGTH = 8192 sentry_sdk.init(project.value, default_integrations=False, release=get_version(), integrations=integrations, traces_sample_rate=1.0, environment=env) sentry_sdk.set_user({"id": dongle_id}) sentry_sdk.set_tag("dirty", is_dirty()) sentry_sdk.set_tag("origin", get_origin()) sentry_sdk.set_tag("branch", get_branch()) sentry_sdk.set_tag("commit", get_commit()) sentry_sdk.set_tag("device", HARDWARE.get_device_type()) if project == SentryProject.SELFDRIVE: sentry_sdk.Hub.current.start_session()
def main(): params = Params() dongle_id = params.get("DongleId").decode('utf-8') cloudlog.bind_global(dongle_id=dongle_id, version=get_version(), dirty=is_dirty()) try: while 1: cloudlog.info("starting athena daemon") proc = Process(name='athenad', target=launcher, args=('selfdrive.athena.athenad', 'athenad')) proc.start() proc.join() cloudlog.event("athenad exited", exitcode=proc.exitcode) time.sleep(5) except Exception: cloudlog.exception("manage_athenad.exception") finally: params.delete(ATHENA_MGR_PID_PARAM)
def manager_init() -> None: # update system time from panda set_time(cloudlog) # save boot log #subprocess.call("./bootlog", cwd=os.path.join(BASEDIR, "selfdrive/loggerd")) params = Params() params.clear_all(ParamKeyType.CLEAR_ON_MANAGER_START) default_params: List[Tuple[str, Union[str, bytes]]] = [ ("CompletedTrainingVersion", "0"), ("DisengageOnAccelerator", "0"), ("HasAcceptedTerms", "0"), ("OpenpilotEnabledToggle", "1"), ("IsMetric", "1"), # HKG ("LateralControl", "TORQUE"), ("UseClusterSpeed", "0"), ("LongControlEnabled", "0"), ("MadModeEnabled", "1"), ("IsLdwsCar", "0"), ("LaneChangeEnabled", "0"), ("AutoLaneChangeEnabled", "0"), ("SccSmootherSlowOnCurves", "0"), ("SccSmootherSyncGasPressed", "0"), ("StockNaviDecelEnabled", "0"), ("KeepSteeringTurnSignals", "0"), ("HapticFeedbackWhenSpeedCamera", "0"), ("DisableOpFcw", "0"), ("ShowDebugUI", "0"), ("NewRadarInterface", "0"), ] if not PC: default_params.append(("LastUpdateTime", datetime.datetime.utcnow().isoformat().encode('utf8'))) if params.get_bool("RecordFrontLock"): params.put_bool("RecordFront", True) if not params.get_bool("DisableRadar_Allow"): params.delete("DisableRadar") # set unset params for k, v in default_params: if params.get(k) is None: params.put(k, v) # is this dashcam? if os.getenv("PASSIVE") is not None: params.put_bool("Passive", bool(int(os.getenv("PASSIVE", "0")))) if params.get("Passive") is None: raise Exception("Passive must be set to continue") # Create folders needed for msgq try: os.mkdir("/dev/shm") except FileExistsError: pass except PermissionError: print("WARNING: failed to make /dev/shm") # set version params params.put("Version", get_version()) params.put("TermsVersion", terms_version) params.put("TrainingVersion", training_version) params.put("GitCommit", get_commit(default="")) params.put("GitBranch", get_short_branch(default="")) params.put("GitRemote", get_origin(default="")) # set dongle id reg_res = register(show_spinner=True) if reg_res: dongle_id = reg_res else: serial = params.get("HardwareSerial") raise Exception(f"Registration failed for device {serial}") os.environ['DONGLE_ID'] = dongle_id # Needed for swaglog if not is_dirty(): os.environ['CLEAN'] = '1' # init logging sentry.init(sentry.SentryProject.SELFDRIVE) cloudlog.bind_global(dongle_id=dongle_id, version=get_version(), dirty=is_dirty(), device=HARDWARE.get_device_type())
def main(): def get_influxdb_line(measurement: str, value: float, timestamp: datetime, tags: dict): res = f"{measurement}" for k, v in tags.items(): res += f",{k}={str(v)}" res += f" value={value} {int(timestamp.timestamp() * 1e9)}\n" return res # open statistics socket ctx = zmq.Context().instance() sock = ctx.socket(zmq.PULL) sock.bind(STATS_SOCKET) # initialize stats directory Path(STATS_DIR).mkdir(parents=True, exist_ok=True) # initialize tags tags = { 'dongleId': Params().get("DongleId", encoding='utf-8'), 'started': False, 'version': get_short_version(), 'branch': get_short_branch(), 'dirty': is_dirty(), 'origin': get_normalized_origin(), 'deviceType': HARDWARE.get_device_type(), } # subscribe to deviceState for started state sm = SubMaster(['deviceState']) last_flush_time = time.monotonic() gauges = {} while True: started_prev = sm['deviceState'].started sm.update() # Update metrics while True: try: metric = sock.recv_string(zmq.NOBLOCK) try: metric_type = metric.split('|')[1] metric_name = metric.split(':')[0] metric_value = metric.split('|')[0].split(':')[1] if metric_type == METRIC_TYPE.GAUGE: gauges[metric_name] = metric_value else: cloudlog.event("unknown metric type", metric_type=metric_type) except Exception: cloudlog.event("malformed metric", metric=metric) except zmq.error.Again: break # flush when started state changes or after FLUSH_TIME_S if (time.monotonic() > last_flush_time + STATS_FLUSH_TIME_S) or (sm['deviceState'].started != started_prev): result = "" current_time = datetime.utcnow().replace(tzinfo=timezone.utc) tags['started'] = sm['deviceState'].started for gauge_key in gauges: result += get_influxdb_line(f"gauge.{gauge_key}", gauges[gauge_key], current_time, tags) # clear intermediate data gauges = {} last_flush_time = time.monotonic() # check that we aren't filling up the drive if len(os.listdir(STATS_DIR)) < STATS_DIR_FILE_LIMIT: if len(result) > 0: stats_path = os.path.join(STATS_DIR, str(int(current_time.timestamp()))) with atomic_write_in_dir(stats_path) as f: f.write(result) else: cloudlog.error("stats dir full")
add_file_handler(cloudlog) cloudlog.error("scons build failed\n" + error_s) # Show TextWindow spinner.close() if not os.getenv("CI"): error_s = "\n \n".join("\n".join(textwrap.wrap(e, 65)) for e in errors) with TextWindow("openpilot failed to build\n \n" + error_s) as t: t.wait_for_exit() exit(1) else: break # enforce max cache size cache_files = [f for f in CACHE_DIR.rglob('*') if f.is_file()] cache_files.sort(key=lambda f: f.stat().st_mtime) cache_size = sum(f.stat().st_size for f in cache_files) for f in cache_files: if cache_size < MAX_CACHE_SIZE: break cache_size -= f.stat().st_size f.unlink() if __name__ == "__main__" and not PREBUILT: spinner = Spinner() spinner.update_progress(0, 100) build(spinner, is_dirty())
def main() -> NoReturn: dongle_id = Params().get("DongleId", encoding='utf-8') def get_influxdb_line(measurement: str, value: Union[float, Dict[str, float]], timestamp: datetime, tags: dict) -> str: res = f"{measurement}" for k, v in tags.items(): res += f",{k}={str(v)}" res += " " if isinstance(value, float): value = {'value': value} for k, v in value.items(): res += f"{k}={v}," res += f"dongle_id=\"{dongle_id}\" {int(timestamp.timestamp() * 1e9)}\n" return res # open statistics socket ctx = zmq.Context().instance() sock = ctx.socket(zmq.PULL) sock.bind(STATS_SOCKET) # initialize stats directory Path(STATS_DIR).mkdir(parents=True, exist_ok=True) # initialize tags tags = { 'started': False, 'version': get_short_version(), 'branch': get_short_branch(), 'dirty': is_dirty(), 'origin': get_normalized_origin(), 'deviceType': HARDWARE.get_device_type(), } # subscribe to deviceState for started state sm = SubMaster(['deviceState']) idx = 0 last_flush_time = time.monotonic() gauges = {} samples: Dict[str, List[float]] = defaultdict(list) while True: started_prev = sm['deviceState'].started sm.update() # Update metrics while True: try: metric = sock.recv_string(zmq.NOBLOCK) try: metric_type = metric.split('|')[1] metric_name = metric.split(':')[0] metric_value = float(metric.split('|')[0].split(':')[1]) if metric_type == METRIC_TYPE.GAUGE: gauges[metric_name] = metric_value elif metric_type == METRIC_TYPE.SAMPLE: samples[metric_name].append(metric_value) else: cloudlog.event("unknown metric type", metric_type=metric_type) except Exception: cloudlog.event("malformed metric", metric=metric) except zmq.error.Again: break # flush when started state changes or after FLUSH_TIME_S if (time.monotonic() > last_flush_time + STATS_FLUSH_TIME_S) or ( sm['deviceState'].started != started_prev): result = "" current_time = datetime.utcnow().replace(tzinfo=timezone.utc) tags['started'] = sm['deviceState'].started for key, value in gauges.items(): result += get_influxdb_line(f"gauge.{key}", value, current_time, tags) for key, values in samples.items(): values.sort() sample_count = len(values) sample_sum = sum(values) stats = { 'count': sample_count, 'min': values[0], 'max': values[-1], 'mean': sample_sum / sample_count, } for percentile in [0.05, 0.5, 0.95]: value = values[int(round(percentile * (sample_count - 1)))] stats[f"p{int(percentile * 100)}"] = value result += get_influxdb_line(f"sample.{key}", stats, current_time, tags) # clear intermediate data gauges.clear() samples.clear() last_flush_time = time.monotonic() # check that we aren't filling up the drive if len(os.listdir(STATS_DIR)) < STATS_DIR_FILE_LIMIT: if len(result) > 0: stats_path = os.path.join( STATS_DIR, f"{current_time.timestamp():.0f}_{idx}") with atomic_write_in_dir(stats_path) as f: f.write(result) idx += 1 else: cloudlog.error("stats dir full")
def manager_init(): # update system time from panda set_time(cloudlog) # save boot log subprocess.call("./bootlog", cwd=os.path.join(BASEDIR, "selfdrive/loggerd")) params = Params() params.clear_all(ParamKeyType.CLEAR_ON_MANAGER_START) default_params = [ ("CompletedTrainingVersion", "0"), ("HasAcceptedTerms", "0"), ("OpenpilotEnabledToggle", "1"), ] if not PC: default_params.append( ("LastUpdateTime", datetime.datetime.utcnow().isoformat().encode('utf8'))) if params.get_bool("RecordFrontLock"): params.put_bool("RecordFront", True) if not params.get_bool("DisableRadar_Allow"): params.delete("DisableRadar") # set unset params for k, v in default_params: if params.get(k) is None: params.put(k, v) # is this dashcam? if os.getenv("PASSIVE") is not None: params.put_bool("Passive", bool(int(os.getenv("PASSIVE")))) if params.get("Passive") is None: raise Exception("Passive must be set to continue") # Create folders needed for msgq try: os.mkdir("/dev/shm") except FileExistsError: pass except PermissionError: print("WARNING: failed to make /dev/shm") # set version params params.put("Version", get_version()) params.put("TermsVersion", terms_version) params.put("TrainingVersion", training_version) params.put("GitCommit", get_commit(default="")) params.put("GitBranch", get_short_branch(default="")) params.put("GitRemote", get_origin(default="")) # set dongle id reg_res = register(show_spinner=True) if reg_res: dongle_id = reg_res else: serial = params.get("HardwareSerial") raise Exception(f"Registration failed for device {serial}") os.environ['DONGLE_ID'] = dongle_id # Needed for swaglog if not is_dirty(): os.environ['CLEAN'] = '1' cloudlog.bind_global(dongle_id=dongle_id, version=get_version(), dirty=is_dirty(), device=HARDWARE.get_device_type()) if is_comma_remote() and not (os.getenv("NOLOG") or os.getenv("NOCRASH") or PC): crash.init() crash.bind_user(id=dongle_id) crash.bind_extra(dirty=is_dirty(), origin=get_origin(), branch=get_short_branch(), commit=get_commit(), device=HARDWARE.get_device_type())