Reputation: 1
Where is actually the output of shuffle & sort phase of a MR programming stored is it stored in local disk or in HDFS?
Upvotes: 0
Views: 394
Reputation: 1882
Of course it stored in local disk because there is no need to duplicate the temporary file. if the map task fail it will retry .
here is the default setting in mapred-site.xml
mapreduce.cluster.local.dir ${hadoop.tmp.dir}/mapred/local
Upvotes: 0
Reputation: 4458
The data is stored in the Local File System (Not HDFS). This is tipically a temporary directory which is cleaned after Hadoop job completes and the configuration is in core-site.xml
.
Upvotes: 1