Hello, 1st of all hadoop needs to use S3 as primary file system. So inside hadoop configuration core-site.xml you need to set fs.default.name with a value of the following form: s3n://your-bucket-name
After this, the way i've done it in hive 0.6 and i assume it still works: alter table my_table add partition (p1=a,p2=b) location "s3n://your-bucket-name/path-to-folder-for-partition" This worked for me without any issues. I assume the other way you provided should work as well, but probably there is an issue with the evaluation of the query... Florin On 11 November 2011 22:11, jiang licht <licht_ji...@yahoo.com> wrote: > Check if this link provides any help: > > http://aws.amazon.com/elasticmapreduce/faqs/#hive-2 > > read " Are there new features in Hive specific to Amazon Elastic > MapReduce?" > > and > > http://aws.amazon.com/articles/2856 > > Best regards, > Michael > ------------------------------ > *From:* Raimon Bosch <raimon.bo...@gmail.com> > *To:* user@hive.apache.org > *Sent:* Friday, November 11, 2011 10:40 AM > *Subject:* load data from s3 to hive > > > Hi, > > I have read that hadoop supports native operations over S3 Filesystem so > you're able to perform operations like: > > hadoop fs -ls s3n://mybucket/my_folder/ > > or: > > hadoop fs -copy s3n://mybucket/my_folder /tmp/my_folder > > I'm wondering why hive is not able to perform similar operations. It would > be a very good feature load data directly from S3 to Hive, something like: > > hive -e "LOAD DATA LOCAL INPATH 's3n://mybucket/my_hive_table' INTO > TABLE my_hive_table PARTITION(dt='2011-11-11');" > > Right now this is not possible. What do you think? Which classes should be > changed? > > > -- Florin
