开源项目lager2-application启动

lager_sup

1.ets表:lager_config:

% 初使化时往表里增加2条数据
1.1 Sink
{lager_event, loglevel} => {Int, []}
例:{lager_event,loglevel}, {255,[]}
其中: Int为8位的整形,即下面的and操作
  -define(DEBUG, 128).
  -define(INFO, 64).
  -define(NOTICE, 32).
  -define(WARNING, 16).
  -define(ERROR, 8).
  -define(CRITICAL, 4).
  -define(ALERT, 2).
  -define(EMERGENCY, 1).
  -define(LOG_NONE, 0).

1.2 Handlers
{_global, handlers} => []

2.三个子进程one_for_one:

{lager, {gen_event, start_link, [{local, lager_event}]},
        permanent, 5000, worker, dynamic}
{lager_handler_watcher_sup, {lager_handler_watcher_sup, start_link, []},
        permanent, 5000, supervisor, [lager_handler_watcher_sup]}
//可选,根据env(crash_log)的值
{lager_crash_log, {lager_crash_log, start_link, [File, MaxBytes,
  RotationSize, RotationDate, RotationCount, RotationMod]},
  permanent, 5000, worker, [lager_crash_log]}

3.lager_event:

% 后面操作中增加事件

4.lager_handler_watcher_sup.erl:

simple_one_for_one:
{lager_handler_watcher, {lager_handler_watcher, start_link, []},
    temporary, 5000, worker, [lager_handler_watcher]}
% 后面通过start_child增加

5.lager_crash_log.erl:

% 后面单独对它分析

lager_app:boot/0

Handler=>lager_backend_throttle

相关参数

{async_threshold, 20},      % 默认undefined
{async_threshold_window, 5} % 默认undefined

a.如没设定async_threshold或值小于0
不执行以下操作

b.如没设定async_threshold_window
async_threshold_window = async_threshold * 0.2(向下取整)

c.两参数不为整数
throw异常

d. 都正常

启动handler=>lager_backend_throttle:

% 启动一个lager_handler_watcher_sup的child: 
supervisor:start_child(lager_handler_watcher_sup, [Sink, lager_backend_throttle, [Threshold, Window]])
其中:
% Sink=lager_event
% Threshold=env(async_threshold)
% Window=env(async_threshold_window)

说明:
会先生成simple_one_for_one的lager_handler_watcher,再启动gen_event
会增加参数{sink, Sink}
Handler=>lager_manager_killer

参数:

{killer_hwm, 1000}              % 默认undefined
{killer_reinstall_after, 5000}  % 默认5000
% This means if the sink's mailbox size exceeds 1000 messages,
% kill the entire sink and reload it after 5000 milliseconds.

% a. 没有配置killer_hwm参数
不执行以下操作

% b.没有配置killer_reinstall_after参数
默认为5000

% c.两参数不是大于0整数
throw异常

% d.一切正常

启动handler=>lager_manager_killer:

% 启动一个lager_handler_watcher_sup的child: 
supervisor:start_child(lager_handler_watcher_sup, [Sink, lager_manager_killer, [HWM, ReinstallTimer]])
其中:
% Sink: lager_event
% HWM: env(killer_hwm)
% ReinstallTimer: env(killer_reinstall_after)

说明:
会先生成simple_one_for_one的lager_handler_watcher,再启动gen_event
配置文件配置handlers

配置文件:

{handlers, [
  {lager_console_backend, [{level, info}]},
  {lager_file_backend, [{file, "error.log"}, {level, error}]},
  {lager_file_backend, [{file, "console.log"}, {level, info}]}
 ]},

% a.没有设备配置文件
不执行以下操作

% b.handlers不是列表
throw异常

% c.一切正常

把配件文件转为格式:{Module, Config},即:

[
  {lager_console_backend, [{level, info}]},
  {
    {lager_file_backend, "error.log"}, 
    [{file, "error.log"}, {level, error}]
  },
  {
    {lager_file_backend, "console.log"}, 
    [{file, "console.log"}, {level, info}]
  }
]

对上面配置文件handlers列表中的每个handler,执行:

lager_app:check_handler_config(Module, Config)
1. 如果是lager_console_backend类型:
  直接返回ok
2. 如果是lager_file_backend类型:
  a. Fs=get(__lager_file_backend_filenames) % 进程字典
    有些字典: 返回值
    无字典: 返回ordsets:new()
  b. put(__lager_file_backend_filenames, ordsets:add_element(F, Fs))
    参数:
    F: "error.log" | "console.log"
    Fs: 进程字典指针

启动handler=>按配置文件启动3个:

start_handler(Sink, Module, Config)
  % Sink: lager_event
  % Module: 
    % lager_console_backend | 
    % {lager_file_backend, "error.log"} | 
    % {lager_file_backend, "console.log"}
  % Config: [{file, "error.log"}, {level, error}]
  % 说明: handlers配置,对list中的每一项启动一个child
  {ok, Watcher} = supervisor:start_child(lager_handler_watcher_sup, [Sink, Module, Config])

往表lager_config中插入数据:

{{_global, handlers} => [{Module, Watcher, Sink} | _OtherList]}
其中:
Module: {lager_file_backend, "error.log"}
Watcher: 上一步start_child的pid
Sink: lager_event
update_loglevel_config模块

获取Traces:

% 后面单独对它分析

get_loglevels(lager_event):

1.获取所有handlers:
  gen_event:which_handlers(Sink)

2.执行call/4
  对名为lager_event的gen_event,对它所有的handlers,执行:
      call/4(request:get_loglevel)

3.合并得到最小log级别,确定什么级别的日志会自动丢弃

4.更新表lager_config中值:

key: {lager_event, loglevel}
value: {MinLog, Traces}
如:
{lager_event,loglevel}, {255,[]}
start_error_logger_handler/3

参数:

{error_logger_redirect, true}   % 默认:true
{error_logger_hwm, 50}          % 默认:0
{error_logger_whitelist, [Logger.ErrorHandler]} %  Stop lager removing Logger's:error_logger handler
{error_logger_groupleader_strategy, undefined}  % 默认为undefined

步骤:

1.如error_logger_redirect => false:
  直接返回[]
2.参数error_logger_groupleader_strategy的值
  2.1 如为: undefined:
    直接返回: handle
  2.2 如为: handle | ignore | mirror
    直接返回: 此值
  2.3 throw 异常

3.supervisor:start_child(lager_handler_watcher_sup, [error_logger, error_logger_lager_h, [HWM, GlStrategy]])
  HWM: 配置error_logger_hwm的值
  GlStrategy: 配置error_logger_groupleader_strategy的值

boot(all_extra)

参数::

extra_sinks
# 后面单独对它分析

boot('traces')

参数::

traces

说明::

1. 需要用到goldrush, 后面单独对它分析
2. traces相关, 后面单独对它分析