Turns out this was a permissions issue.  We had Write and Execute permissions, but not Read.  As a result, the Archive Remover failed (threw an NPE which logback hides unless debug="true" is set in the config file), which threw off the rollover code (didn't complete properly), which threw off the internal Output Streams, cycling between 3 different but faulty states…

 

I've submitted a (low priority) usability request that could help others track permissions issues like this down…

 

Bob Sandiford | Principal Engineer SirsiDynix

P: 800.288.8020 X6943 | Bob.Sandiford@sirsidynix.com

www.sirsidynix.com

 

Join the conversation: Like us on Facebook! Follow us on Twitter!

 

From: Logback-user [mailto:logback-user-bounces@qos.ch] On Behalf Of Bob Sandiford
Sent: Tuesday, May 13, 2014 4:38 PM
To: logback-user@qos.ch
Subject: [logback-user] logback skipping log files on AWS EC2

 

I'm using logback for logging from an app deployed in Tomcat, with a fairly simple setup (see code fragments). We use a RollingFileAppender, with TimeBasedRollingPolicy, set for daily rollover. When running locally, everything appears to be fine. When running in AWS in an EC2 instance, I'm seeing that some log files are missing.

I wrote a really simple app that does nothing but log once per second with a counter, and then a logback config that rolls every minute. For this particular test, we're seeing every third log file is missing.

So, for example, we'll get:

-rw-r--r-- 1 tomcat tomcat  891 May 13 18:46 logtest_tomcat.2014-05-13_1845.0.log.gz

-rw-r--r-- 1 tomcat tomcat  499 May 13 18:47 logtest_tomcat.2014-05-13_1846.0.log.gz

-rw-r--r-- 1 tomcat tomcat  541 May 13 18:49 logtest_tomcat.2014-05-13_1848.0.log.gz

-rw-r--r-- 1 tomcat tomcat  519 May 13 18:50 logtest_tomcat.2014-05-13_1849.0.log.gz

-rw-r--r-- 1 tomcat tomcat  532 May 13 18:52 logtest_tomcat.2014-05-13_1851.0.log.gz

-rw-r--r-- 1 tomcat tomcat  510 May 13 18:53 logtest_tomcat.2014-05-13_1852.0.log.gz

-rw-r--r-- 1 tomcat tomcat  536 May 13 18:55 logtest_tomcat.2014-05-13_1854.0.log.gz

-rw-r--r-- 1 tomcat tomcat 1226 May 13 18:56 logtest_tomcat.2014-05-13_1855.0.log.gz

-rw-r--r-- 1 tomcat tomcat  531 May 13 18:58 logtest_tomcat.2014-05-13_1857.0.log.gz

-rw-r--r-- 1 tomcat tomcat  496 May 13 18:59 logtest_tomcat.2014-05-13_1858.0.log.gz

-rw-r--r-- 1 tomcat tomcat 1244 May 13 19:01 logtest_tomcat.2014-05-13_1900.0.log.gz

-rw-r--r-- 1 tomcat tomcat  496 May 13 19:02 logtest_tomcat.2014-05-13_1901.0.log.gz

-rw-r--r-- 1 tomcat tomcat  514 May 13 19:04 logtest_tomcat.2014-05-13_1903.0.log.gz

-rw-r--r-- 1 tomcat tomcat  500 May 13 19:05 logtest_tomcat.2014-05-13_1904.0.log.gz

-rw-r--r-- 1 tomcat tomcat  522 May 13 19:07 logtest_tomcat.2014-05-13_1906.0.log.gz

The file format is yyyy-mm-dd_HHmm - so you can see that 1847, 1850, 1853, 1856, 1859, 1902, 1905 are all missing.

I've checked the contents - the sequential numbering on the log statements jumps by 60 for the missing logs - so it's not that multiple minutes are being collapsed into a single rolled over log.

We also thought it might be due to our Splunk forwarder - we ran the test both with and without the Splunk forwarder running, and got the same results - every third log file is missing.

Here's the logback appender for this test:

<appender name="daily" class="ch.qos.logback.core.rolling.RollingFileAppender">

    <file>${bc.logs.home}/logtest_tomcat.log</file>

    <rollingPolicy class="ch.qos.logback.core.rolling.TimeBasedRollingPolicy">

        <!-- Rollover every minute for this test -->

        <fileNamePattern>${bc.logs.home}/logtest_tomcat.%d{yyyy-MM-dd_HHmm}.%i.log.gz</fileNamePattern>

        <timeBasedFileNamingAndTriggeringPolicy class="ch.qos.logback.core.rolling.SizeAndTimeBasedFNATP">

            <!-- or whenever the file size reaches 250MB -->

            <maxFileSize>250MB</maxFileSize>

        </timeBasedFileNamingAndTriggeringPolicy>

        <maxHistory>60</maxHistory>

    </rollingPolicy>

    <append>true</append>

    <encoder>

        <pattern>%d{"yyyy-MM-dd HH:mm:ss,SSS z", UTC} [%thread] %-5level %logger{36} - %msg%n</pattern>

    </encoder>

</appender>

And here's my little driver class (the 'BCLog' is a simple wrapper around slf4j logging, instantiated by

        Logger log = LoggerFactory.getLogger(clazz);

)

package com.sirsidynix.logtest.biz.svc.impl;

 

import com.sirsidynix.bccommon.util.BCLog;

import com.sirsidynix.bccommon.util.BCLogFactory;

import org.springframework.beans.factory.DisposableBean;

import org.springframework.beans.factory.InitializingBean;

 

public class JustLogIt implements InitializingBean, DisposableBean

{

    private static final BCLog LOG = BCLogFactory.getLog(JustLogIt.class);

 

    private Thread thread;

 

    @Override

    public void afterPropertiesSet() throws Exception

    {

        LOG.info("Firing up JustLogIt thread");

        thread = new Thread(){

            @Override

            public void run()

            {

                long iteration = 0;

                while (true)

                {

                    try

                    {

                        Thread.sleep(1000);

                        iteration++;

                        LOG.info("Logging Iteration " + iteration);

                    }

                    catch (InterruptedException e)

                    {

                        LOG.info("LogIt thread sleep interrupted!!!");

                    }

                }

            }

        };

        thread.start();

    }

 

 

    @Override

    public void destroy() throws Exception

    {

        LOG.info("Shutting down JustLogIt thread");

        thread.interrupt();

    }

}

Any ideas?

Thanks!

 

 

Bob Sandiford | Principal Engineer SirsiDynix

P: 800.288.8020 X6943 | Bob.Sandiford@sirsidynix.com

www.sirsidynix.com

 

Join the conversation: Like us on Facebook! Follow us on Twitter!