1. 简介
在此 Codelab 中,您将使用 gRPC 创建一个客户端和服务器,它们将构成一个用 Python 编写的路线映射应用的基础。
完成本教程后,您将拥有一个简单的 gRPC HelloWorld 应用,该应用已通过 gRPC OpenTelemetry 插件进行插桩,并且您将能够在 Prometheus 中看到导出的可观测性指标。
学习内容
- 如何为现有的 gRPC Python 应用设置 OpenTelemetry 插件
- 运行本地 Prometheus 实例
- 将指标导出到 Prometheus
- 查看 Prometheus 信息中心内的指标
2. 准备工作
所需条件
- git
- curl
- build-essential
- Python 3.9 或更高版本。如需查看针对具体平台的 Python 安装说明,请参阅 Python 设置和使用。或者,使用 uv 或 pyenv 等工具安装非系统 Python。
- pip 9.0.1 版或更高版本。
- venv 来创建 Python 虚拟环境。
安装必备项:
sudo apt-get update -y
sudo apt-get upgrade -y
sudo apt-get install -y git curl build-essential clang
sudo apt install python3
sudo apt install python3-pip python3-venv
获取代码
为了简化学习过程,此 Codelab 提供了预构建的源代码框架,可帮助您快速入门。以下步骤将指导您在应用中对 gRPC OpenTelemetry 插件进行插桩。
grpc-codelab
此 Codelab 的框架源代码位于此 GitHub 目录中。如果您不想自行实现代码,可以在 completed 目录中找到已完成的源代码。
首先,克隆 grpc Codelab 代码库,然后 cd 进入 grpc-python-opentelemetry 文件夹:
git clone https://github.com/grpc-ecosystem/grpc-codelabs.git
cd grpc-codelabs/codelabs/grpc-python-opentelemetry/
或者,您也可以下载仅包含 Codelab 目录的 .zip 文件,然后手动将其解压缩。
我们先创建一个新的 Python 虚拟环境 (venv),以将项目的依赖项与系统软件包隔离开来:
python3 -m venv --upgrade-deps .venv
在 bash/zsh shell 中激活虚拟环境:
source .venv/bin/activate
对于 Windows 和非标准 shell,请参阅 https://docs.python.org/3/library/venv.html#how-venvs-work 中的表格。
接下来,使用以下命令在环境中安装依赖项:
python -m pip install -r requirements.txt
3. 注册 OpenTelemetry 插件
我们需要一个 gRPC 应用来添加 gRPC OpenTelemetry 插件。在此 Codelab 中,我们将使用简单的 gRPC HelloWorld 客户端和服务器,并使用 gRPC OpenTelemetry 插件对其进行插桩处理。
第一步是在客户端中注册配置了 Prometheus 导出器的 OpenTelemetry 插件。使用您惯用的编辑器打开 start_here/observability_greeter_client.py。首先,添加相关依赖项和宏,如下所示 -
import logging
import time
import grpc
import grpc_observability
import helloworld_pb2
import helloworld_pb2_grpc
from opentelemetry.exporter.prometheus import PrometheusMetricReader
from opentelemetry.sdk.metrics import MeterProvider
from prometheus_client import start_http_server
_SERVER_PORT = "50051"
_PROMETHEUS_PORT = 9465
然后,转换 run(),使其如下所示 -
def run():
# Start Prometheus client
start_http_server(port=_PROMETHEUS_PORT, addr="0.0.0.0")
meter_provider = MeterProvider(metric_readers=[PrometheusMetricReader()])
otel_plugin = grpc_observability.OpenTelemetryPlugin(
meter_provider=meter_provider
)
otel_plugin.register_global()
with grpc.insecure_channel(target=f"localhost:{_SERVER_PORT}") as channel:
stub = helloworld_pb2_grpc.GreeterStub(channel)
# Continuously send RPCs every second.
while True:
try:
response = stub.SayHello(helloworld_pb2.HelloRequest(name="You"))
print(f"Greeter client received: {response.message}")
time.sleep(1)
except grpc.RpcError as rpc_error:
print("Call failed with code: ", rpc_error.code())
# Deregister is not called in this example, but this is required to clean up.
otel_plugin.deregister_global()
下一步是将 OpenTelemetry 插件添加到服务器。打开 start_here/observability_greeter_server.py 并添加相关依赖项和宏,使其看起来像这样 -
from concurrent import futures
import logging
import time
import grpc
import grpc_observability
import helloworld_pb2
import helloworld_pb2_grpc
from opentelemetry.sdk.metrics import MeterProvider
from opentelemetry.exporter.prometheus import PrometheusMetricReader
from prometheus_client import start_http_server
_SERVER_PORT = "50051"
_PROMETHEUS_PORT = 9464
然后,转换 run(),使其如下所示 -
def serve():
# Start Prometheus client
start_http_server(port=_PROMETHEUS_PORT, addr="0.0.0.0")
meter_provider = MeterProvider(metric_readers=[PrometheusMetricReader()])
otel_plugin = grpc_observability.OpenTelemetryPlugin(
meter_provider=meter_provider
)
otel_plugin.register_global()
server = grpc.server(
thread_pool=futures.ThreadPoolExecutor(max_workers=10),
)
helloworld_pb2_grpc.add_GreeterServicer_to_server(Greeter(), server)
server.add_insecure_port("[::]:" + _SERVER_PORT)
server.start()
print("Server started, listening on " + _SERVER_PORT)
server.wait_for_termination()
# Deregister is not called in this example, but this is required to clean up.
otel_plugin.deregister_global()
4. 运行示例并查看指标
如需运行服务器,请运行以下命令:
cd start_here
python -m observability_greeter_server
如果设置成功,您将看到以下服务器输出 -
Server started, listening on 50051
在服务器运行时,在另一个终端上运行客户端 -
# Run the below commands to cd to the working directory and activate virtual environment in the new terminal
cd grpc-codelabs/codelabs/grpc-python-opentelemetry/
source .venv/bin/activate
cd start_here
python -m observability_greeter_client
成功运行后,输出结果如下所示:
Greeter client received: Hello You
Greeter client received: Hello You
Greeter client received: Hello You
由于我们已设置 gRPC OpenTelemetry 插件以使用 Prometheus 导出指标。这些指标将分别在 localhost:9464(服务器)和 localhost:9465(客户端)上提供。
如需查看客户端指标,请执行以下操作:
curl localhost:9465/metrics
结果的格式应为:
# HELP python_gc_objects_collected_total Objects collected during gc
# TYPE python_gc_objects_collected_total counter
python_gc_objects_collected_total{generation="0"} 241.0
python_gc_objects_collected_total{generation="1"} 163.0
python_gc_objects_collected_total{generation="2"} 0.0
# HELP python_gc_objects_uncollectable_total Uncollectable objects found during GC
# TYPE python_gc_objects_uncollectable_total counter
python_gc_objects_uncollectable_total{generation="0"} 0.0
python_gc_objects_uncollectable_total{generation="1"} 0.0
python_gc_objects_uncollectable_total{generation="2"} 0.0
# HELP python_gc_collections_total Number of times this generation was collected
# TYPE python_gc_collections_total counter
python_gc_collections_total{generation="0"} 78.0
python_gc_collections_total{generation="1"} 7.0
python_gc_collections_total{generation="2"} 0.0
# HELP python_info Python platform information
# TYPE python_info gauge
python_info{implementation="CPython",major="3",minor="10",patchlevel="9",version="3.10.9"} 1.0
# HELP process_virtual_memory_bytes Virtual memory size in bytes.
# TYPE process_virtual_memory_bytes gauge
process_virtual_memory_bytes 1.868988416e+09
# HELP process_resident_memory_bytes Resident memory size in bytes.
# TYPE process_resident_memory_bytes gauge
process_resident_memory_bytes 4.1680896e+07
# TYPE process_resident_memory_bytes gauge 21:20:16 [154/966]
process_resident_memory_bytes 4.1680896e+07
# HELP process_start_time_seconds Start time of the process since unix epoch in seconds.
# TYPE process_start_time_seconds gauge
process_start_time_seconds 1.72375679833e+09
# HELP process_cpu_seconds_total Total user and system CPU time spent in seconds.
# TYPE process_cpu_seconds_total counter
process_cpu_seconds_total 0.38
# HELP process_open_fds Number of open file descriptors.
# TYPE process_open_fds gauge
process_open_fds 9.0
# HELP process_max_fds Maximum number of open file descriptors.
# TYPE process_max_fds gauge
process_max_fds 4096.0
# HELP target_info Target metadata
# TYPE target_info gauge
target_info{service_name="unknown_service",telemetry_sdk_language="python",telemetry_sdk_name="opentelemetry",telemetry_sdk_version="1.26.0"} 1.0
# HELP grpc_client_attempt_started_total Number of client call attempts started
# TYPE grpc_client_attempt_started_total counter
grpc_client_attempt_started_total{grpc_method="other",grpc_target="localhost:50051"} 18.0
# HELP grpc_client_attempt_sent_total_compressed_message_size_bytes Compressed message bytes sent per client call attempt
# TYPE grpc_client_attempt_sent_total_compressed_message_size_bytes histogram
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="0.0"} 0.0
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="5.0"} 18.0
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="10.0"} 18.0
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="25.0"} 18.0
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="50.0"} 18.0
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="75.0"} 18.0
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="100.0"} 18.0
grpc_client_attempt_sent_total_compressed_message_size_bytes_bucket{grpc_method="other",grpc_status="OK",grpc_target="localhost:50051",le="250.0"} 18.0
同样,对于服务器端指标 -
curl localhost:9464/metrics
5. 在 Prometheus 上查看指标
在此示例中,我们将设置一个 Prometheus 实例,该实例将抓取使用 Prometheus 导出指标的 gRPC 示例客户端和服务器。
使用给定的链接下载适用于您的平台的最新版 Prometheus,或使用以下命令:
curl -sLO https://github.com/prometheus/prometheus/releases/download/v3.7.3/prometheus-3.7.3.linux-amd64.tar.gz
然后使用以下命令提取并运行该文件:
tar xvfz prometheus-*.tar.gz
cd prometheus-*
创建一个包含以下内容的 Prometheus 配置文件:
cat > grpc_otel_python_prometheus.yml <<EOF
scrape_configs:
- job_name: "prometheus"
scrape_interval: 5s
static_configs:
- targets: ["localhost:9090"]
- job_name: "grpc-otel-python"
scrape_interval: 5s
static_configs:
- targets: ["localhost:9464", "localhost:9465"]
EOF
使用新配置启动 Prometheus -
./prometheus --config.file=grpc_otel_python_prometheus.yml
此命令会将客户端和服务器 Codelab 进程的指标配置为每 5 秒抓取一次。
前往 http://localhost:9090/graph 查看指标。例如,以下查询:
histogram_quantile(0.5, rate(grpc_client_attempt_duration_seconds_bucket[1m]))
将显示一个图表,其中包含使用 1 分钟作为分位数计算的时间窗口的尝试延迟时间中位数。
查询率 -
increase(grpc_client_attempt_duration_seconds_bucket[1m])
6. (可选)用户练习
在 Prometheus 信息中心内,您会发现 QPS 较低。看看您能否在示例中找到一些限制 QPS 的可疑代码。
对于热衷于此的开发者,客户端代码限制为在给定时刻只能有一个待处理的 RPC。您可以修改此设置,以便客户端发送更多 RPC,而无需等待之前的 RPC 完成。(此问题的解决方案尚未提供。)