[ https://issues.apache.org/jira/browse/METRON-1185?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16165565#comment-16165565 ]
ASF GitHub Bot commented on METRON-1185: ---------------------------------------- Github user cestella commented on a diff in the pull request: https://github.com/apache/metron/pull/755#discussion_r138776403 --- Diff: metron-stellar/stellar-common/src/main/scripts/stellar --- @@ -28,7 +28,8 @@ elif [ -e /usr/lib/bigtop-utils/bigtop-detect-javahome ]; then . /usr/lib/bigtop-utils/bigtop-detect-javahome fi -export HBASE_CONFIGS=/etc/hbase/conf +export HBASE_CONFIGS=$(hbase classpath) export METRON_VERSION=${project.version} export METRON_HOME=/usr/metron/$METRON_VERSION -java $JVMFLAGS -cp "$HBASE_CONFIGS:$METRON_HOME/lib/*" org.apache.metron.stellar.stellar.shell.StellarShell "$@" +export STELLAR_LIB=$(find $METRON_HOME/lib/ -name metron-enrichment*.jar) --- End diff -- It will pick up the profiler functions, they're included in the enrichment jar (that's actually what motivated this, `PROFILE_GET` didn't work randomly based on the order of the classpath). We probably should include the management functions directly, though. I'll adjust. > Stellar REPL does not work on a kerberized cluster when calling functions > interacting with HBase > ------------------------------------------------------------------------------------------------ > > Key: METRON-1185 > URL: https://issues.apache.org/jira/browse/METRON-1185 > Project: Metron > Issue Type: Improvement > Reporter: Casey Stella > > Because we put all the jars on the classpath from $METRON_HOME/lib via a > glob, if we get the wrong order, we get classpath issues around Stopwatch > (i.e. more recent guava may sneak onto the classpath first). This is > non-deterministic and can cause issues randomly. Instead of a glob, we > should choose one of the uber jars with all the metron Stellar functions > baked in and allow users to provide any extra directories we should include > via an env variable. Also, we are not pulling the full hbase classpath, so > in a kerberized cluster we do not get all of the configs on the classpath, so > we can't find always find tgts. -- This message was sent by Atlassian JIRA (v6.4.14#64029)