For some circumstances I need to force flushing in logback’s file appender immediately. I’ve found in docs this option is enabled by default. Mysteriously this doesn’t work. As I see in the sources underlying process involves BufferedOutputSream correctly. Is there any issues with BufferedOutputSream.flush() ? Probably this is rather related to the flushing issue.
Update:
I found the issue on Windows XP Pro SP 3 and on Red Hat Enterprise Linux Server release 5.3 (Tikanga).
I used these libs:
jcl-over-slf4j-1.6.6.jar
logback-classic-1.0.6.jar
logback-core-1.0.6.jar
slf4j-api-1.6.6.jar
The logback.xml is:
<configuration>
<appender name="FILE" class="ch.qos.logback.core.rolling.RollingFileAppender">
<file>/somepath/file.log</file>
<rollingPolicy class="ch.qos.logback.core.rolling.FixedWindowRollingPolicy">
<fileNamePattern>file.log.%i</fileNamePattern>
<minIndex>1</minIndex>
<maxIndex>3</maxIndex>
</rollingPolicy>
<triggeringPolicy class="ch.qos.logback.core.rolling.SizeBasedTriggeringPolicy">
<maxFileSize>5MB</maxFileSize>
</triggeringPolicy>
<encoder>
<pattern>%d{HH:mm:ss.SSS} [%thread] %-5level %logger - %msg%n</pattern>
</encoder>
</appender>
<root level="debug">
<appender-ref ref="FILE"/>
</root>
</configuration>
Updated:
I’d provide a unit test but that seems not so simple.
Let me describe the issue more clearly.
- Event of logging occurred
- Event is passed into file appender
- Event is serialized with defined pattern
- Serialized message of event is passed to the file appender and is
about to write out to output stream - Writing to the stream is finished, output stream is flushed (I’ve
checked the implementation). Note thatimmidiateFlushis true by
default so methodflush()is invoked explicitly - No result in the file!
A bit later when some underlying buffer was flowed the event appears in the file.
So the question is: does output stream guarantee immediate flush?
To be honest I’ve already solve this by implementing my own ImmediateRollingFileAppender that leverages facility of FileDescriptor of immediate syncing. Anybody interested in can follow this.
So this is not a logback issue.
I decided to bring my solution to everybody.
Let me clarify first of all that this is not a logback issue and not a JRE problem. This is described in the javadoc and generally shouldn’t be an issue until you are faced with some old school integration solutions over the file syncing.
So this is a logback appender implemented to flush immediately:
This is corresponding output stream utility class. Because of some methods and fields of original
ResilientOutputStreamBasethat was supposed for extending initially have packaged access modifiers I had to extendOutputStreaminstead and just copy the rest and unchanged ofResilientOutputStreamBaseandResilientFileOutputStreamto this new one. I just display the changed code:And finally the config: