-
问题
表现形式:在定时任务的常驻进程(死循环)里 yii 配置的log 文件日志无法存储。
在带有终止(exit , die ) 的调试里日志写入正常。
启动循环后(进程常驻)时,无法写入日志。
-
分析
分析源码后得出,日志写入过程是先写入内存,程序终止时用函数
register_shutdown_function 监测到后,调用flush 函数 将日志消息存入指定日志。
-
解决
手动触发 flush
Yii::$app->log->getLogger()->flush(true);
-
后遗症
若有使用自定义文件存储 app.log 和 自定义路径 都会存储
-
部分源码
<?php
/**
* @link http://www.yiiframework.com/
* @copyright Copyright (c) 2008 Yii Software LLC
* @license http://www.yiiframework.com/license/
*/
namespace yii\log;
use Yii;
use yii\base\Component;
/**
* Logger records logged messages in memory and sends them to different targets if [[dispatcher]] is set.
*
* A Logger instance can be accessed via `Yii::getLogger()`. You can call the method [[log()]] to record a single log message.
* For convenience, a set of shortcut methods are provided for logging messages of various severity levels
* via the [[Yii]] class:
*
* - [[Yii::trace()]]
* - [[Yii::error()]]
* - [[Yii::warning()]]
* - [[Yii::info()]]
* - [[Yii::beginProfile()]]
* - [[Yii::endProfile()]]
*
* For more details and usage information on Logger, see the [guide article on logging](guide:runtime-logging).
*
* When the application ends or [[flushInterval]] is reached, Logger will call [[flush()]]
* to send logged messages to different log targets, such as [[FileTarget|file]], [[EmailTarget|email]],
* or [[DbTarget|database]], with the help of the [[dispatcher]].
*
* @property array $dbProfiling The first element indicates the number of SQL statements executed, and the
* second element the total time spent in SQL execution. This property is read-only.
* @property float $elapsedTime The total elapsed time in seconds for current request. This property is
* read-only.
* @property array $profiling The profiling results. Each element is an array consisting of these elements:
* `info`, `category`, `timestamp`, `trace`, `level`, `duration`, `memory`, `memoryDiff`. The `memory` and
* `memoryDiff` values are available since version 2.0.11. This property is read-only.
*
* @author Qiang Xue <qiang.xue@gmail.com>
* @since 2.0
*/
class Logger extends Component
{
/**
* Error message level. An error message is one that indicates the abnormal termination of the
* application and may require developer's handling.
*/
const LEVEL_ERROR = 0x01;
/**
* Warning message level. A warning message is one that indicates some abnormal happens but
* the application is able to continue to run. Developers should pay attention to this message.
*/
const LEVEL_WARNING = 0x02;
/**
* Informational message level. An informational message is one that includes certain information
* for developers to review.
*/
const LEVEL_INFO = 0x04;
/**
* Tracing message level. An tracing message is one that reveals the code execution flow.
*/
const LEVEL_TRACE = 0x08;
/**
* Profiling message level. This indicates the message is for profiling purpose.
*/
const LEVEL_PROFILE = 0x40;
/**
* Profiling message level. This indicates the message is for profiling purpose. It marks the
* beginning of a profiling block.
*/
const LEVEL_PROFILE_BEGIN = 0x50;
/**
* Profiling message level. This indicates the message is for profiling purpose. It marks the
* end of a profiling block.
*/
const LEVEL_PROFILE_END = 0x60;
/**
* @var array logged messages. This property is managed by [[log()]] and [[flush()]].
* Each log message is of the following structure:
*
* ```
* [
* [0] => message (mixed, can be a string or some complex data, such as an exception object)
* [1] => level (integer)
* [2] => category (string)
* [3] => timestamp (float, obtained by microtime(true))
* [4] => traces (array, debug backtrace, contains the application code call stacks)
* [5] => memory usage in bytes (int, obtained by memory_get_usage()), available since version 2.0.11.
* ]
* ```
*/
public $messages = [];
/**
* @var int how many messages should be logged before they are flushed from memory and sent to targets.
* Defaults to 1000, meaning the [[flush]] method will be invoked once every 1000 messages logged.
* Set this property to be 0 if you don't want to flush messages until the application terminates.
* This property mainly affects how much memory will be taken by the logged messages.
* A smaller value means less memory, but will increase the execution time due to the overhead of [[flush()]].
*/
public $flushInterval = 1000;
/**
* @var int how much call stack information (file name and line number) should be logged for each message.
* If it is greater than 0, at most that number of call stacks will be logged. Note that only application
* call stacks are counted.
*/
public $traceLevel = 0;
/**
* @var Dispatcher the message dispatcher
*/
public $dispatcher;
/**
* Initializes the logger by registering [[flush()]] as a shutdown function.
*/
public function init()
{
parent::init();
register_shutdown_function(function () {
// make regular flush before other shutdown functions, which allows session data collection and so on
$this->flush();
// make sure log entries written by shutdown functions are also flushed
// ensure "flush()" is called last when there are multiple shutdown functions
register_shutdown_function([$this, 'flush'], true);
});
}
/**
* Logs a message with the given type and category.
* If [[traceLevel]] is greater than 0, additional call stack information about
* the application code will be logged as well.
* @param string|array $message the message to be logged. This can be a simple string or a more
* complex data structure that will be handled by a [[Target|log target]].
* @param int $level the level of the message. This must be one of the following:
* `Logger::LEVEL_ERROR`, `Logger::LEVEL_WARNING`, `Logger::LEVEL_INFO`, `Logger::LEVEL_TRACE`,
* `Logger::LEVEL_PROFILE_BEGIN`, `Logger::LEVEL_PROFILE_END`.
* @param string $category the category of the message.
*/
public function log($message, $level, $category = 'application')
{
$time = microtime(true);
$traces = [];
if ($this->traceLevel > 0) {
$count = 0;
$ts = debug_backtrace(DEBUG_BACKTRACE_IGNORE_ARGS);
array_pop($ts); // remove the last trace since it would be the entry script, not very useful
foreach ($ts as $trace) {
if (isset($trace['file'], $trace['line']) && strpos($trace['file'], YII2_PATH) !== 0) {
unset($trace['object'], $trace['args']);
$traces[] = $trace;
if (++$count >= $this->traceLevel) {
break;
}
}
}
}
$this->messages[] = [$message, $level, $category, $time, $traces, memory_get_usage()];
if ($this->flushInterval > 0 && count($this->messages) >= $this->flushInterval) {
$this->flush();
}
}
/**
* Flushes log messages from memory to targets.
* @param bool $final whether this is a final call during a request.
*/
public function flush($final = false)
{
$messages = $this->messages;
// https://github.com/yiisoft/yii2/issues/5619
// new messages could be logged while the existing ones are being handled by targets
$this->messages = [];
if ($this->dispatcher instanceof Dispatcher) {
$this->dispatcher->dispatch($messages, $final);
}
}
.....
其他部分已省略
本文探讨了在Yii框架中,当应用运行在死循环(常驻进程)模式下,日志记录功能出现的问题及解决方案。具体表现为日志无法正常写入文件。通过分析发现,Yii的日志机制会在程序终止时自动刷新日志缓存。文章提供了手动触发flush函数的方法来解决该问题。
1141

被折叠的 条评论
为什么被折叠?



