Hi all, I was recently going through the https://drill.apache.org/docs/ documentation to setup a Distributed Mode Drill cluster that will consume data off of HDFS. Our cluster runs Ubuntu 14.04. I found that following the PAM installation steps at https://drill.apache.org/docs/configuring-user-authentication/ had an unfortunate side effect. It appears that setting
export DRILLBIT_JAVA_OPTS="-Djava.library.path=/opt/pam/" in drill-env.sh as the documentation suggests causes the java library path to be overwritten, knocking off important files such as libhadoop.so The net effect of this is that when performing certain actions, such as querying a snappy compressed file, the relevant library files would not be available causing the action to fail. In the case of querying snappy files, this would result in the error: 2016-06-17 09:36:29,385 [USER-rpc-event-queue] ERROR o.a.d.exec.server.rest.QueryWrapper - Query Failed org.apache.drill.common.exceptions.UserRemoteException: DATA_READ ERROR: Failure reading JSON file - native snappy library not available: this version of libhadoop was built without snappy support. To resolve this issue, I instead added the libjpam.so file to drill-env.sh by using: LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/opt/pam/" This appears to have resolved the issue so far, allowing me to both authenticate with pam and query snappy compressed files. I'm emailing then with two purposes: * Should I be concerned that this diversion from the documentation will cause issues in the future? * To let others know, in case they run into odd behavior after installing jpam, that this may help. Thanks, Ben
