Boost Spring Boot Performance: Master Asynchronous Logback Configuration

Introduction

Today we discuss Logback asynchronous logging configuration, covering three goals: output logs to files and separate them by LEVEL, reduce disk I/O with asynchronous logging, and explain the principle behind asynchronous log output.

01. Configuration file logback-spring.xml

Spring Boot includes Logback and SLF4J by default, so only the configuration file needs to be created. Logback automatically loads a file named logback-spring.xml or logback.xml from the classpath.

The recommended setup separates ERROR logs from other logs and rolls files daily:

<?xml version="1.0" encoding="utf-8"?>
<configuration>
    <property resource="logback.properties"/>
    <appender name="CONSOLE-LOG" class="ch.qos.logback.core.ConsoleAppender">
        <layout class="ch.qos.logback.classic.PatternLayout">
            <pattern>[%d{yyyy-MM-dd HH:mm:ss.SSS}] [%C] [%t] [%L] [%-5p] %m%n</pattern>
        </layout>
    </appender>
    <!-- INFO and below (excluding ERROR) -->
    <appender name="INFO-LOG" class="ch.qos.logback.core.rolling.RollingFileAppender">
        <filter class="ch.qos.logback.classic.filter.LevelFilter">
            <level>ERROR</level>
            <onMatch>DENY</onMatch>
            <onMismatch>ACCEPT</onMismatch>
        </filter>
        <encoder>
            <pattern>[%d{yyyy-MM-dd HH:mm:ss.SSS}] [%C] [%t] [%L] [%-5p] %m%n</pattern>
        </encoder>
        <rollingPolicy class="ch.qos.logback.core.rolling.TimeBasedRollingPolicy">
            <fileNamePattern>${LOG_INFO_HOME}/%d.log</fileNamePattern>
            <maxHistory>30</maxHistory>
        </rollingPolicy>
    </appender>
    <appender name="ERROR-LOG" class="ch.qos.logback.core.rolling.RollingFileAppender">
        <filter class="ch.qos.logback.classic.filter.ThresholdFilter">
            <level>ERROR</level>
        </filter>
        <encoder>
            <pattern>[%d{yyyy-MM-dd HH:mm:ss.SSS}] [%C] [%t] [%L] [%-5p] %m%n</pattern>
        </encoder>
        <rollingPolicy class="ch.qos.logback.core.rolling.TimeBasedRollingPolicy">
            <fileNamePattern>${LOG_ERROR_HOME}/%d.log</fileNamePattern>
            <maxHistory>30</maxHistory>
        </rollingPolicy>
    </appender>
    <root level="info">
        <appender-ref ref="CONSOLE-LOG"/>
        <appender-ref ref="INFO-LOG"/>
        <appender-ref ref="ERROR-LOG"/>
    </root>
</configuration>

Key tags explained: <appender>: defines where and how logs are written (console or file). <filter>: selects log levels to include or exclude. <encoder>: formats the log output. <rollingPolicy>: controls log file rotation, e.g., daily rolling with a 30‑day retention.

02. Asynchronous Log Output Feature

Switching from synchronous to asynchronous logging removes the blocking disk I/O from the request thread. Add an AsyncAppender that references the existing file appenders:

<!-- Asynchronous output -->
<appender name="ASYNC-INFO" class="ch.qos.logback.classic.AsyncAppender">
    <discardingThreshold>0</discardingThreshold>
    <queueSize>256</queueSize>
    <appender-ref ref="INFO-LOG"/>
</appender>

<appender name="ASYNC-ERROR" class="ch.qos.logback.classic.AsyncAppender">
    <discardingThreshold>0</discardingThreshold>
    <queueSize>256</queueSize>
    <appender-ref ref="ERROR-LOG"/>
</appender>

03. Performance Test

Hardware: 6‑core CPU, 8 GB RAM. Test tool: Apache JMeter.

Synchronous logging : 100 threads, ramp‑up 0 s, TPS = 44.2 /sec.

Asynchronous logging : same load, TPS = 497.5 /sec, more than ten times faster.

Asynchronous Logging Principle

The process starts at Logger.info() and reaches AsyncAppenderBase.append():

protected void append(E eventObject) {
    if (!this.isQueueBelowDiscardingThreshold() || !this.isDiscardable(eventObject)) {
        this.preprocess(eventObject);
        this.put(eventObject);
    }
}

If the queue is not over the discarding threshold, the event is placed into a blocking ArrayBlockingQueue (default size 256, configurable). A dedicated worker thread then consumes events:

E e = parent.blockingQueue.take();
aai.appendLoopOnAppenders(e);

The worker thread invokes appendLoopOnAppenders to forward the log event to the underlying file appenders. The core methods involved in writing are write and encode.

Original article: https://juejin.cn/post/6844903909920604174