[ https://issues.apache.org/jira/browse/KAFKA-4142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15531931#comment-15531931 ]
Umesh Chaudhary commented on KAFKA-4142: ---------------------------------------- [~cmhillerman], Did you compare the size of both files which you see the date modified changed but are having same name? >From the doc I see: the log is cleaned by recopying each log segment but omitting any key that appears in the offset map with a higher offset than what is found in the segment (i.e. messages with a key that appears in the dirty section of the log). I believe when you will compare the size of both files, you will find some messages are cleaned based on retention policy. > Log files in /data dir date modified keeps being updated? > --------------------------------------------------------- > > Key: KAFKA-4142 > URL: https://issues.apache.org/jira/browse/KAFKA-4142 > Project: Kafka > Issue Type: Bug > Affects Versions: 0.10.0.0 > Environment: CentOS release 6.8 (Final) > uname -a > Linux 2.6.32-642.1.1.el6.x86_64 #1 SMP Tue May 31 21:57:07 UTC 2016 x86_64 > x86_64 x86_64 GNU/Linux > Reporter: Clint Hillerman > Priority: Minor > > The date modified of the kafka logs (the main ones specified by logs.dirs in > the config) keep getting updated and set to the exact same time. > For example: > Say I had two log and index files ( date modified - file name): > 20160901:10:00:01 - 0001.log > 20160901:10:00:01 -0001.index > 20160902:10:00:01 -0002.log > 20160902:10:00:01 -0002.index > Later I notice the logs are getting way to old for the retention time. I then > go look at the log dir and I see this: > 20160903:10:00:01 - 0001.log > 2016090310:00:01 -0001.index > 20160903:10:00:01 -0002.log > 20160903:10:00:01 -0002.index > 20160903:10:00:01 -0003.log > 20160903:10:00:01 -0003.index > 20160904:10:00:01 -0004.log > 20160904:10:00:01 -0004.index > The first two log files had there date modified moved forward for some > reason. They were updated from 0901 and 0902 to 0903. > It seems to happen periodically. The new logs that kafka writes out have the > correct time stamp. > This causes the logs to not be deleted. Right now I just touch the log files > to an older date and they are deleted right away. > Any help would be appreciated. Also, I'll explain the problem better if this > doesn't make sense. > Thanks, -- This message was sent by Atlassian JIRA (v6.3.4#6332)