[ https://issues.apache.org/jira/browse/HADOOP-1288?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12542401 ]
Enis Soztutar commented on HADOOP-1288: --------------------------------------- bq. wonder whether it has to do with the default value of 12 hours for mapred.userlog.retain.hours. Yes, indeed it has. When a new task is started the task cleans the logs in the log directory if modifiedTime < now - mapred.userlog.retain.hours. We can fix this by first checking if the task owning that log file is still running, or we may just increase default value and mark this as wont fix, what do others think? > missing userlogs > ---------------- > > Key: HADOOP-1288 > URL: https://issues.apache.org/jira/browse/HADOOP-1288 > Project: Hadoop > Issue Type: Bug > Components: mapred > Affects Versions: 0.12.3 > Reporter: Christian Kunz > > I noticed that a long-running job taking more than 1 day can have reduce > tasks with missing stderr and stdout log directories (the syslog files exist). > I wonder whether it has to do with the default value of 12 hours for > mapred.userlog.retain.hours, resulting in deletion of the stderr/stdout > directories for reduce tasks idling till they start to actually reduce. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.