Regarding Hadoop Log Retention for Hadoop 0.20.2

2011-12-07 Thread Sahana Bhat
Hi, We use Hadoop 0.20.2 version.The log4j.properties file has a property *hadoop.tasklog.logsRetainHours *(mentioned as 24 hours by default) * *which we have set to 12.Despite this property being set we still have old userlogs in the system.Any idea why the attempt directories (that have

Re: Regarding Hadoop Log Retention for Hadoop 0.20.2

2011-12-07 Thread Harsh J
Sahana, IIRC, mapred.userlog.retain.hours (24h default) controls this in my environment and it seems to work fine on my cluster. Are you sure you have tasklogs older than 24h lying around? It might even be a bug that may have been fixed in the subsequent 0.20 releases that went out recently. On

Re: Regarding Hadoop Log Retention for Hadoop 0.20.2

2011-12-07 Thread Sahana Bhat
Hi Harsh, Thanks for the reply. I realized that the property you mentioned was missing in my mapred-site.xml. I added the entry and it works just fine. Was my assumption that *hadoop.tasklog.logsRetainHours *in log4j.properties will do the same wrong? What is this property for in that

Re: Regarding Hadoop Log Retention for Hadoop 0.20.2

2011-12-07 Thread Vinod Kumar Vavilapalli
That is a stale property in log4j.properties that isn't used for anything. There are other such things like noKeepSplits, purgeLogSplits etc which seem to be stale too. mapred.userlog.retain.hours should work just fine as Harsh correctly mentioned. HTH, +Vinod On Dec 7, 2011, at 2:48 AM,

Re: Regarding Hadoop Log Retention for Hadoop 0.20.2

2011-12-07 Thread Sahana Bhat
Hi Vinod, Thanks for that info :). Regards, Sahana On Thu, Dec 8, 2011 at 3:30 AM, Vinod Kumar Vavilapalli vino...@hortonworks.com wrote: That is a stale property in log4j.properties that isn't used for anything. There are other such things like noKeepSplits, purgeLogSplits etc which