Hi qingyang, 1. You do not need to install shark on every node. 2. Not really sure..it's just a warning so I'd see if it works despite it 3. You need to provide the actual hdfs path, e.g. hdfs://namenode/user2/vols.csv, see this thread https://groups.google.com/forum/#!topic/tachyon-users/3Da4zcHKBbY
Lastly as your questions are more shark than spark related there is a separate shark user group that might be more helpful. Hope this helps On Thu, Mar 6, 2014 at 3:25 AM, qingyang li <liqingyang1...@gmail.com>wrote: > just a addition for #3, i have such configuration in shark-env.sh: > ---- > export HADOOP_HOME=/usr/lib/hadoop > export HADOOP_CONF_DIR=/etc/hadoop/conf > export HIVE_HOME=/usr/lib/hive/ > #export HIVE_CONF_DIR=/etc/hive/conf > export MASTER=spark://bigdata001:7077 > ----- > > > 2014-03-06 16:20 GMT+08:00 qingyang li <liqingyang1...@gmail.com>: > > hi, spark community, i have setup 3 nodes cluster using spark 0.9 and >> shark 0.9, My question is : >> 1. is there any neccessary to install shark on every node since it is a >> client to use spark service ? >> 2. when i run shark-withinfo, i got such warning: >> WARN shark.SharkEnv: Hive Hadoop shims detected local mode, but Shark is >> not running locally. >> WARN shark.SharkEnv: Setting mapred.job.tracker to 'Spark_1394093746930' >> (was 'local') >> what does this log want to tell us ? >> is it a problem to run shark? >> 3. i want to load data from hdfs , so i run "LOAD DATA INPATH >> '/user/root/input/test.txt' into table b; " , but i got this error:No files >> matching path file:/user/root/input/test.txt , but this file exists on >> hdfs. >> >> thanks. >> > >