[ 
https://issues.apache.org/jira/browse/TIKA-4050?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17743175#comment-17743175
 ] 

Josh Burchard commented on TIKA-4050:
-------------------------------------

Hey Tim. Thanks for the response. I poked around at what you mentioned, but 
never quite grokked it. The -Dcf  jvm option was specifying the dtikacfg.xml 
file, while -Dll specifies the log4jTika.xml file.  ${env:tika.server.id} only 
spits out "Domino" since that's what I'd set in my config file, so was I 
missing a way to get a unique ID?

Anyhow, I came up with a workaround by changing Log4j2's RollingFile fileName 
attribute to have this value instead:  
fileName="${sys:ll}.${date:yyMMddHHmmssSSSS}".  This gets me a new unique 
logfile for every child process.  It'd be better if log4j2 supported retrieving 
the # of seconds since epoch but, alas, they don't.

 

> Forked child logs are overwritten upon restart
> ----------------------------------------------
>
>                 Key: TIKA-4050
>                 URL: https://issues.apache.org/jira/browse/TIKA-4050
>             Project: Tika
>          Issue Type: Bug
>          Components: core
>    Affects Versions: 2.8.0
>         Environment: Windows 10 Enterprise
>            Reporter: Josh Burchard
>            Priority: Major
>         Attachments: dtikacfg.xml, log4j2-async.xml, log4jTika.xml
>
>
> In my config file I specify a log file path for the forked child to use.  A 
> problem arises when the child has a problem and the watchdog restarts it 
> because the new child overwrites the previous child's log so I lose the 
> details about what the child was doing prior to a restart.  Requesting that 
> logging is preserved for all children while the watchdog is resident.
>  
> Reproduction steps:
> [^dtikacfg.xml] [^log4jTika.xml]
>  # ^Launch tika server:^
> ^java.exe -Djt="c:\tikatmp0000" -Dll="c:\tika0000.log" 
> -Dcf="c:\log4jTika.xml" -jar "c:\tika-server-standard-2.8.0.jar" -c 
> "c:\dtikacfg.xml"^
>  # ^Peek at the tika0000.log file and take note of the timestamps.^
>  # ^Kill the Tika child task and allow the watchdog to restart it.^
>  # ^Note that the previous content has been overwritten by log lines with new 
> timestamps.^
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to