【腾讯犀牛鸟开源课题实战】prometheus插件专项建设（PUSH模式支持等） #175

Heaven2024 · 2024-08-26T02:58:08Z

issue
完成了push模式以及promethues的鉴权支持

note:
修正了md文档相关问题
另外prometheus示例文件中的运行脚本无法直接运行，编译后的路径名和config参数不对：
/home/TRPC/trpc-cpp/examples/features/prometheus/run.sh: line 9: ./bazel-bin/examples/features/prometheus/client/client_config: No such file or directory

ERROR: unknown command line flag 'config'

Prometheus鉴权

weimch · 2024-10-11T09:57:39Z

这个问题有修复吗？

unknown command line flag 'config'

weimch · 2024-10-11T09:58:07Z

.bazelrc

@@ -3,3 +3,5 @@ build --copt=-O2
 #build --copt=-g --strip=never
 build --jobs 16
 #test --cache_test_results=no --test_output=errors
+
+build --define trpc_include_prometheus=true


默认关闭prometheus，这行可以删掉

Heaven2024 · 2024-10-11T10:44:04Z

这个问题有修复吗？

unknown command line flag 'config'

fixed

weimch · 2024-10-11T11:10:57Z

examples/features/prometheus/proxy/push.cc

@@ -0,0 +1,22 @@
+#include <chrono>


这个文件和框架无关，没必要增加，用法放在文档就好了

weimch · 2024-10-11T11:11:25Z

examples/features/prometheus/proxy/forward_service.cc

@@ -77,6 +77,12 @@ ::trpc::Status ForwardServiceImpl::Route(::trpc::ServerContextPtr context,
      "counter_name", "counter_desc", {{"const_counter_key", "const_counter_value"}});
  ::prometheus::Counter& counter = counter_family->Add({{"counter_key", "counter_value"}});
  counter.Increment(random_num);
+
+  if (::trpc::prometheus::PushMetricsInfo()) {


为啥这里还需要手动调用呢？不能配置一下yaml文件就生效吗？

weimch · 2024-10-11T11:32:00Z

profiling/sysvars.data

@@ -0,0 +1,78 @@
+<div id="layer1">


去掉不需要的文件

已删除。

weimch · 2024-10-11T11:33:48Z

examples/features/prometheus/proxy/BUILD

@@ -44,3 +44,13 @@ cc_library(
        "@trpc_cpp//trpc/metrics/prometheus:prometheus_metrics_api",
    ],
 )
+
+cc_binary(


不需要push这个文件，去掉与之相关的编译引入

weimch · 2024-10-11T11:35:28Z

trpc/admin/prometheus_handler.cc

+  } else {
+    TRPC_LOG_INFO("can not found prometheus auth config");
+    has_cfg = false;
+  }


构造函数做了太复杂的事情，可以定义一个Init函数，把这部分逻辑放在Init函数里

已修改。

weimch · 2024-10-11T11:38:13Z

trpc/admin/prometheus_handler.cc


 void PrometheusHandler::CommandHandle(http::HttpRequestPtr req, rapidjson::Value& result,
                                      rapidjson::Document::AllocatorType& alloc) {
  static std::unique_ptr<::prometheus::Serializer> serializer = std::make_unique<::prometheus::TextSerializer>();
+
+  if (has_cfg) {


has_cfg变量名不好，看起来只跟鉴权判断相关，同时可以考虑把鉴权相关信息统一放在一个结构体里。

已修改。

weimch · 2024-10-11T11:40:32Z

trpc/admin/prometheus_handler.cc


 void PrometheusHandler::CommandHandle(http::HttpRequestPtr req, rapidjson::Value& result,
                                      rapidjson::Document::AllocatorType& alloc) {
  static std::unique_ptr<::prometheus::Serializer> serializer = std::make_unique<::prometheus::TextSerializer>();
+
+  if (has_cfg) {
+    std::string token = req->GetHeader("Authorization");


这块逻辑有啥用？看起来只是判断用户名和密码是否匹配，判断之后有啥用？

只有用户名密码都正确的情况下，才会返回metric数据，否则拒绝请求。

用户名和密码，在prometheus的gateway服务哪里能配置呢？文档有给出吗？

想起来了，这是pull模式的，那用户名和密码在prometheus服务器里哪里能配置呢？

文档我还没有写，确认一下，Prometheus鉴权相关的使用方法是直接添加在prometheus_metrics.md吗？

trpc/admin/prometheus_handler.cc

weimch · 2024-10-11T11:42:49Z

trpc/admin/prometheus_handler.h

@@ -26,6 +33,12 @@ class PrometheusHandler : public AdminHandlerBase {

  void CommandHandle(http::HttpRequestPtr req, rapidjson::Value& result,
                     rapidjson::Document::AllocatorType& alloc) override;
+ private:
+  PrometheusConfig prometheus_conf_;   


还是不理解鉴权相关的参数放admin服务的意图，可以描述下

因为Prometheus拉取数据是要走admin服务的，我感觉只有在这里才能拿到http包头中的用户名密码信息，才能进行鉴权。

看实现，只需要填充username和password就好了吧？不需要保留prometheus_conf_，只需要填充CommandHandle里的username和password字段

已修改。

weimch · 2024-10-11T11:43:53Z

trpc/common/plugin.h

@@ -71,7 +71,7 @@ class Plugin : public RefCounted<Plugin> {

  /// @brief Stop the runtime environment of the plugin
  virtual void Stop() noexcept {}
-
+  


记得用clang-format把所有代码文件都格式化一遍（使用项目根目录的.clang-format配置的格式化规范）

weimch · 2024-10-11T11:44:16Z

trpc/common/plugin.h

@@ -71,7 +71,7 @@ class Plugin : public RefCounted<Plugin> {

  /// @brief Stop the runtime environment of the plugin
  virtual void Stop() noexcept {}
-
+  


这里出现了不必要的空格

weimch · 2024-10-12T06:57:52Z

trpc/metrics/prometheus/prometheus_conf.h

+    bool enabled = false;
+    std::string gateway_url;
+    std::string job_name;
+    int push_interval_seconds = 15;


命名为 push_interval，单位为毫秒吧

weimch · 2024-10-12T06:58:11Z

trpc/metrics/prometheus/prometheus_conf.h

+  struct PushMode {
+    bool enabled = false;
+    std::string gateway_url;
+    std::string job_name;


job_name这个字段有什么用？

github-actions · 2024-10-12T07:08:05Z

CLA Assistant Lite bot All contributors have signed the CLA ✍️ ✅

weimch · 2024-10-15T02:39:51Z

examples/features/prometheus/proxy/trpc_cpp_fiber.yaml

@@ -44,6 +44,11 @@ plugins:
      const_labels:
        const_key1: const_value1
        const_key2: const_value2
+      push_mode:


例子需要演示pull模式和push模式，应该给出2个文件配置

weimch · 2024-10-15T03:02:10Z

trpc/admin/prometheus_handler.cc

+        "Failed to obtain Prometheus plugin configuration from the framework configuration file. Default configuration "
+        "will be used.");
+  }
+  Init();


构造函数别执行太复杂的事情，Init不能放在外面执行吗？

已修改。

weimch · 2024-10-15T06:30:24Z

trpc/admin/prometheus_handler.cc


 void PrometheusHandler::CommandHandle(http::HttpRequestPtr req, rapidjson::Value& result,
                                      rapidjson::Document::AllocatorType& alloc) {
  static std::unique_ptr<::prometheus::Serializer> serializer = std::make_unique<::prometheus::TextSerializer>();

+  if (auth_conf_.username.size() && auth_conf_.password.size()) {
+    std::string token = req->GetHeader("Authorization");
+    auto splited = Split(token, ' ');


鉴权部分单独提出一个类私有成员接口

已修改。

weimch · 2024-10-15T06:30:57Z

trpc/admin/prometheus_handler.cc

+    auto sp = Split(username_pwd, ':');
+    if (sp.size() != 2) {
+      result.AddMember("message", "wrong request without authorization", alloc);
+      TRPC_LOG_INFO("error token: " << token);


使用错误日志宏 TRPC_FMT_ERROR

已修改。

weimch · 2024-10-15T06:32:41Z

trpc/admin/prometheus_handler.cc

+    }
+
+    auto username = sp[0];
+    if (username != auth_conf_.username) {


用户名和密码应该一起判断，如果不对，统一返回 wrong username or password 的信息，现在这种实现，攻击者能猜对用户名

已修改。

weimch · 2024-10-15T06:34:22Z

trpc/admin/prometheus_handler.h

@@ -16,6 +16,12 @@

 #include "trpc/admin/admin_handler.h"
 #include "trpc/util/prometheus.h"
+#include "trpc/util/http/base64.h"


用clang-format格式化一下，头文件顺序需要按照字母序顺序排列

已修改。

weimch · 2024-10-15T06:36:37Z

trpc/admin/prometheus_handler.h

+  PrometheusConfig prometheus_conf_;   
+
+  struct AuthConf {
+    std::string username;


为啥不使用token的方式来鉴权呢？

是可以使用token，一开始的实现也是token，但是我查了资料，pushgateway没办法用token来鉴权。所以如果pull模式用token的话，就相当于是两套鉴权模式了。

push和pull确实是两套鉴权模式吧，配置区分开就好

weimch · 2024-10-15T06:54:24Z

trpc/metrics/prometheus/BUILD

@@ -73,15 +83,20 @@ cc_library(
        ":prometheus_conf",
        ":prometheus_conf_parser",
        "//trpc/util:prometheus",
+        "@com_github_jupp0r_prometheus_cpp//core",
+


不必要的换行，BUILD文件使用 buildifier 格式化一下

weimch · 2024-10-15T07:02:54Z

trpc/metrics/prometheus/prometheus_metrics.cc


 namespace trpc {

+// PrometheusMetrics::~PrometheusMetrics() {


去掉无用注释

weimch · 2024-10-15T07:05:24Z

trpc/metrics/prometheus/prometheus_pusher.h

+    bool Push(const std::vector<::prometheus::MetricFamily>& metrics);
+    bool AsyncPush(const std::vector<::prometheus::MetricFamily>& metrics);
+
+    void registerCollectable(const std::weak_ptr<::prometheus::Collectable>& collectable) {


接口命名风格，方法名首字母大写，驼峰式命名

weimch · 2024-10-15T07:06:13Z

trpc/metrics/prometheus/prometheus_pusher.cc

+
+PrometheusPusher::PrometheusPusher(const std::string& gateway_url, const std::string& job_name)
+    : job_name_(job_name) {
+    gateway_ = std::make_unique<::prometheus::Gateway>("localhost","9091", job_name);


这里没有使用配置

weimch · 2024-10-15T07:07:33Z

trpc/metrics/prometheus/prometheus_pusher.cc

+bool PrometheusPusher::Push(const std::vector<::prometheus::MetricFamily>& metrics)
+ {
+    int push_result = gateway_->Push();
+    if (push_result != 200) {


魔数，应该使用Prometheus里专门的错误码

weimch · 2024-10-15T07:08:13Z

trpc/metrics/prometheus/prometheus_pusher.cc

+    if (push_result != 200) {
+        std::ostringstream oss;
+        oss << "Failed to push metrics to Pushgateway. Error code: " << push_result;
+        TRPC_LOG_ERROR(oss.str());


这样写不行吗？

TRPC_LOG_ERROR("Failed to push metrics to Pushgateway. Error code: " << push_result);

weimch · 2024-10-15T07:47:05Z

trpc/metrics/prometheus/prometheus_pusher.cc

+
+    // 等待异步操作完成并获取结果
+    try {
+        int push_result = push_future.get();


这里不是阻塞等吗？从效果上和同步调用有啥区别，不应该用future-then的回调吗

weimch · 2024-10-15T07:47:50Z

trpc/workspace.bzl

@@ -80,6 +80,19 @@ def trpc_workspace(path_prefix = "", repo_name = "", **kwargs):
            urls = zlib_urls,
        )

+
+    http_archive(
+    name = "com_github_curl",


用 bazel 的 buildifier 工具格式化一下

weimch · 2024-10-15T07:48:01Z

trpc/workspace.bzl

@@ -80,6 +80,19 @@ def trpc_workspace(path_prefix = "", repo_name = "", **kwargs):
            urls = zlib_urls,
        )

+
+    http_archive(
+    name = "com_github_curl",


这里为啥要引入curl?

Heaven2024 and others added 8 commits August 26, 2024 02:44

Update prometheus documentation and running scripts

809dfe0

Update prometheus documentation and running scripts

ccfc856

Fixed prometheus documentation and running scripts

8a1de0e

Fixed prometheus documentation and running scripts

c879973

v1

e1916f8

Feature: Add Prometheus pull mode authentication.

b980723

Merge branch 'prometheus' into branch_prometheus_basic

7f42fbd

Merge pull request #1 from leolin49/branch_prometheus_basic

cd24612

Prometheus鉴权

Heaven2024 changed the title ~~Fixed prometheus documentation and running scripts~~ 【腾讯犀牛鸟开源课题实战】prometheus插件专项建设（PUSH模式支持等） Oct 7, 2024

fixed

816bc28

weimch reviewed Oct 11, 2024

View reviewed changes

trpc/admin/prometheus_handler.cc Outdated Show resolved Hide resolved

weimch reviewed Oct 11, 2024

View reviewed changes

weimch reviewed Oct 12, 2024

View reviewed changes

Fixed: Prometheus pull mode authentication.

3d907d6

leolin49 force-pushed the prometheus branch from fc4f149 to 3d907d6 Compare October 12, 2024 10:24

weimch reviewed Oct 15, 2024

View reviewed changes

trpc/metrics/prometheus/prometheus_metrics.cc

namespace trpc {

// PrometheusMetrics::~PrometheusMetrics() {

Copy link

Contributor

weimch Oct 15, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

去掉无用注释

weimch reviewed Oct 15, 2024

View reviewed changes

leolin49 added 2 commits October 16, 2024 20:56

Fixed: Prometheus pull mode authentication.

2309769

Fixed: Prometheus pull mode authentication.

1a5bb30

		@@ -71,7 +71,7 @@ class Plugin : public RefCounted<Plugin> {

		/// @brief Stop the runtime environment of the plugin
		virtual void Stop() noexcept {}


		namespace trpc {

		// PrometheusMetrics::~PrometheusMetrics() {

【腾讯犀牛鸟开源课题实战】prometheus插件专项建设（PUSH模式支持等） #175

Are you sure you want to change the base?

【腾讯犀牛鸟开源课题实战】prometheus插件专项建设（PUSH模式支持等） #175

Conversation

Heaven2024 commented Aug 26, 2024 • edited Loading

weimch commented Oct 11, 2024

Choose a reason for hiding this comment

Heaven2024 commented Oct 11, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

weimch Oct 15, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

github-actions bot commented Oct 12, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

leolin49 Oct 16, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

weimch Oct 15, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Heaven2024 commented Aug 26, 2024 •

edited

Loading

weimch Oct 15, 2024 •

edited

Loading

github-actions bot commented Oct 12, 2024 •

edited

Loading

leolin49 Oct 16, 2024 •

edited

Loading

weimch Oct 15, 2024 •

edited

Loading