feat(core): Read and save system information to tracer (#670)

**Others**
- Fix the problem of repeated startup of start ModelWorker in cluster
deploying.
- Docker image use `Asia/Shanghai` as timezone when LANGUAGE is "zh"

.dockerignore

@@ -1,2 +1,5 @@
 models/
 plugins/
+pilot/data
+pilot/message
+logs/

docker/base/Dockerfile

@@ -3,7 +3,7 @@ ARG BASE_IMAGE="nvidia/cuda:11.8.0-runtime-ubuntu22.04"
 FROM ${BASE_IMAGE}
 ARG BASE_IMAGE
 
-RUN apt-get update && apt-get install -y git python3 pip wget sqlite3 \
+RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y git python3 pip wget sqlite3 tzdata \
     && apt-get clean
 
 ARG BUILD_LOCAL_CODE="false"
@@ -44,11 +44,6 @@ ARG BUILD_LOCAL_CODE="false"
 # COPY the rest of the app
 COPY . /app
 
-# TODO：Need to find a better way to determine whether to build docker image with local code.
-RUN (if [ "${BUILD_LOCAL_CODE}" = "true" ]; \
-    then rm -rf /app/logs && rm -rf /app/pilot/data && rm -rf /app/pilot/message; \
-    fi;)
-
 ARG LOAD_EXAMPLES="true"
 
 RUN (if [ "${LOAD_EXAMPLES}" = "true" ]; \
@@ -57,6 +52,11 @@ RUN (if [ "${LOAD_EXAMPLES}" = "true" ]; \
     && sqlite3 /app/pilot/data/default_sqlite.db < /app/docker/examples/sqls/test_case_info_sqlite.sql; \
     fi;)
 
+RUN (if [ "${LANGUAGE}" = "zh" ]; \
+    then ln -sf /usr/share/zoneinfo/Asia/Shanghai /etc/localtime \
+    && echo "Asia/Shanghai" > /etc/timezone; \
+    fi;)
+
 ENV PYTHONPATH "/app:$PYTHONPATH"
 EXPOSE 5000
 

pilot/model/cluster/embedding/loader.py

@@ -5,6 +5,7 @@
 from pilot.model.parameter import BaseEmbeddingModelParameters
 from pilot.utils.parameter_utils import _get_dict_from_obj
 from pilot.utils.tracer import root_tracer, SpanType, SpanTypeRunName
+from pilot.utils.system_utils import get_system_info
 
 if TYPE_CHECKING:
     from langchain.embeddings.base import Embeddings
@@ -21,6 +22,7 @@ def load(
             "model_name": model_name,
             "run_service": SpanTypeRunName.EMBEDDING_MODEL.value,
             "params": _get_dict_from_obj(param),
+            "sys_infos": _get_dict_from_obj(get_system_info()),
         }
         with root_tracer.start_span(
             "EmbeddingLoader.load", span_type=SpanType.RUN, metadata=metadata

pilot/model/cluster/worker/default_worker.py

@@ -11,6 +11,7 @@
 from pilot.utils.model_utils import _clear_model_cache
 from pilot.utils.parameter_utils import EnvArgumentParser, _get_dict_from_obj
 from pilot.utils.tracer import root_tracer, SpanType, SpanTypeRunName
+from pilot.utils.system_utils import get_system_info
 
 logger = logging.getLogger(__name__)
 
@@ -102,6 +103,7 @@ def start(
             "llm_adapter": str(self.llm_adapter),
             "run_service": SpanTypeRunName.MODEL_WORKER,
             "params": _get_dict_from_obj(model_params),
+            "sys_infos": _get_dict_from_obj(get_system_info()),
         }
         with root_tracer.start_span(
             "DefaultModelWorker.start", span_type=SpanType.RUN, metadata=metadata

pilot/model/cluster/worker/manager.py

@@ -40,6 +40,7 @@
 )
 from pilot.utils.utils import setup_logging
 from pilot.utils.tracer import initialize_tracer, root_tracer, SpanType, SpanTypeRunName
+from pilot.utils.system_utils import get_system_info
 
 logger = logging.getLogger(__name__)
 
@@ -838,6 +839,7 @@ def _start_local_worker(
         metadata={
             "run_service": SpanTypeRunName.WORKER_MANAGER,
             "params": _get_dict_from_obj(worker_params),
+            "sys_infos": _get_dict_from_obj(get_system_info()),
         },
     ):
         worker = _build_worker(worker_params)
@@ -974,6 +976,7 @@ def run_worker_manager(
         os.path.join(LOGDIR, "dbgpt_model_worker_manager_tracer.jsonl"),
         root_operation_name="DB-GPT-WorkerManager-Entry",
     )
+
     _start_local_worker(worker_manager, worker_params)
     _start_local_embedding_worker(
         worker_manager, embedding_model_name, embedding_model_path
@@ -985,11 +988,13 @@ def run_worker_manager(
     if not embedded_mod:
         import uvicorn
 
-        loop = asyncio.get_event_loop()
-        loop.run_until_complete(worker_manager.start())
         uvicorn.run(
             app, host=worker_params.host, port=worker_params.port, log_level="info"
         )
+    else:
+        # Embedded mod, start worker manager
+        loop = asyncio.get_event_loop()
+        loop.run_until_complete(worker_manager.start())
 
 
 if __name__ == "__main__":

pilot/server/dbgpt_server.py

@@ -5,7 +5,6 @@
 
 ROOT_PATH = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 sys.path.append(ROOT_PATH)
-import signal
 from pilot.configs.config import Config
 from pilot.configs.model_config import LLM_MODEL_CONFIG, EMBEDDING_MODEL_CONFIG, LOGDIR
 from pilot.component import SystemApp
@@ -40,6 +39,7 @@
 )
 from pilot.utils.tracer import root_tracer, initialize_tracer, SpanType, SpanTypeRunName
 from pilot.utils.parameter_utils import _get_dict_from_obj
+from pilot.utils.system_utils import get_system_info
 
 static_file_path = os.path.join(os.getcwd(), "server/static")
 
@@ -190,6 +190,7 @@ def run_webserver(param: WebWerverParameters = None):
         metadata={
             "run_service": SpanTypeRunName.WEBSERVER,
             "params": _get_dict_from_obj(param),
+            "sys_infos": _get_dict_from_obj(get_system_info()),
         },
     ):
         param = initialize_app(param)