本地App Engine向Datastore注册实体时遇连接错误

尝试向Datastore写入实体或执行查询时均触发错误，应用未连接到Datastore模拟器。

错误详情 #

调用<function make_call..rpc_call at 0x7fba88531510>时超出3次重试上限，最后异常为503 无法连接所有地址；最后错误：UNKNOWN: ipv4:127.0.0.1:8000: Trying to connect an http1.x server

运行方式：通过python3 main.py启动FastAPI应用，调用创建实体接口/test时触发错误。
期望：在本地Datastore模拟器中保存实体并创建对应的Kind。

相关代码 #

main.py #

import uvicorn
from fastapi import FastAPI
from google.cloud import ndb

app = FastAPI()
ds_client = ndb.Client()

class Neighborhood(ndb.Model):
    db_id = ndb.ComputedProperty(lambda self: self.key.id())
    name = ndb.StringProperty(required=True)
    zone = ndb.StringProperty()
    active = ndb.BooleanProperty(default=True)

@app.get("/test2")
def fetch_neighborhood():
    'get most recent Neighborhood'
    with ds_client.context():
        return Neighborhood.query().fetch()

@app.post("/test")
def store_neighborhood():
    'create new Neighborhood entity in Datastore'
    with ds_client.context():
        namespace = 'client2'
        neighborhood = Neighborhood(id=123, namespace=namespace)
        neighborhood.name = "test"
        neighborhood.put()
    
if __name__ == "__main__":
    uvicorn.run("main:app", host="0.0.0.0", port=8080, reload=True)

启动脚本（bash） #

#!/bin/bash

HOST=localhost
PY_PATH=python27=/usr/bin/python2.7,python3=/usr/bin/python3

export APPLICATION_ID=dev~None
export GOOGLE_APPLICATION_CREDENTIALS="./app_engine_service_account_dev.json"
export DEPLOY_ENVIRONMENT=development
export HTTP_HOST="localhost:8080"

python3 ~/google-cloud-sdk/bin/dev_appserver.py \
    --port=8080 \
    --host=$HOST \
    --support_datastore_emulator=true \
    --runtime_python_path=$PY_PATH \
app-dev.yaml
# --clear_datastore=yes \

app-dev.yaml #

runtime: python310
entrypoint: gunicorn -w 4 -k uvicorn.workers.UvicornWorker main:app
instance_class: F1
automatic_scaling:
  target_cpu_utilization: 0.7
  min_instances: 1
  max_instances: 1
  min_pending_latency: 30ms  # default value
  max_pending_latency: automatic
  max_concurrent_requests: 32
  max_idle_instances: 0

skip_files:
- ^(.*/)?#.*#$
- ^(.*/)?.*~$
- ^(.*/)?.*\.py[co]$
- ^(.*/)?.*/RCS/.*$
- ^(.*/)?\..*$
- ^(.*/)?.*\.bak$
- ^(.*/)?.*deploy.sh$
- ^(.*/)?docs/*$
- ^(.*/)?venv/*$
- ^(.*/)?downloads/*$

env_variables:
  DEPLOY_ENVIRONMENT: 'development'

  DATASTORE_EMULATOR_HOST: 'localhost:8000'
  DATASTORE_EMULATOR_HOST_PATH: 'localhost:8000/datastore'
  DATASTORE_HOST: 'http://localhost:8000'
  HTTP_HOST: 'localhost:8080'

  STORAGE_EMULATOR_HOST: 'http://localhost:9023'

解决方案 #

1. 确保Datastore模拟器已正常启动 #

• 单独启动模拟器：在终端执行以下命令启动Datastore模拟器

  gcloud beta emulators datastore start --host-port=localhost:8000
  

• 若使用dev_appserver.py启动，需检查启动日志，确认模拟器已成功启动（日志中会包含Datastore emulator running at localhost:8000相关信息）

2. 运行应用时加载必要环境变量 #

直接执行python3 main.py时，不会自动加载app-dev.yaml或bash脚本中的环境变量，需手动设置：

# 先加载模拟器环境变量（启动模拟器的终端执行后，复制输出的export命令到当前终端，或直接执行）
$(gcloud beta emulators datastore env-init)
# 设置项目ID（与脚本中一致）
export APPLICATION_ID="dev~None"
# 启动应用
python3 main.py

3. 代码优化：显式指定项目ID #

修改main.py中的NDB客户端初始化，避免自动检测项目ID失败：

ds_client = ndb.Client(project="dev~None")

4. 检查端口占用 #

确认8000端口未被其他程序占用：

lsof -i :8000

若有占用进程，执行kill -9 <进程ID>杀掉后重启模拟器。

内容的提问来源于stack exchange，提问作者Matheus Theu