HDP 2.0 Directory structure confusion
I just followed installation manuals of HDP 2.0 for manual installation using and RPM packages. I got a bit confused regarding LOG and PID directories as those are necessary to define and create e.g. HDFS_LOG_DIR=”/var/log/hadoop/hdfs”; But in reality rpm itself creates directories line /var/hadoop-hdfs, /var/hadoop-mapreduce. /var/hadoop-yarn . More over when I run hadoop it sometimes logs to created directories, sometime in rpm created. I found there is env variable overriding in conf shell scripts e.g. yarn-env.sh which overloads some variables to those from rpm.
I used companion files suggested in manual and followed all process as described on fresh machine – no collision with previous versions avoided.
Can someone clarify usage of those directories? To me it seems more as a bug so I would like to avoid introduction of other problems when fixing this.