[ https://issues.apache.org/jira/browse/HDFS-13124?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16372593#comment-16372593 ]
Jianfei Jiang commented on HDFS-13124: -------------------------------------- Also meet this problem in 3.0. In the code of trunk, the sleep time seems to have been changed to 1s, but the balancer is still too quick to get a ps result. The related code is in {{haddop-functions.sh#hadoop_start_daemon_wrapper}}. {code:java} sleep 1 # capture the ulimit output ulimit -a >> "${outfile}" 2>&1 # shellcheck disable=SC2086 if ! ps -p $! >/dev/null 2>&1; then return 1 fi return 0 {code} > hadoop-daemon.sh exits with 1 when running HDFS balancer on balanced cluster > ---------------------------------------------------------------------------- > > Key: HDFS-13124 > URL: https://issues.apache.org/jira/browse/HDFS-13124 > Project: Hadoop HDFS > Issue Type: Bug > Components: balancer & mover, scripts > Affects Versions: 2.7.3 > Reporter: Zbigniew Kostrzewa > Priority: Minor > > When running HDFS balancer via {{sbin/start-balancer.sh}} script on a > balanced cluster the script exits with 1 though the CLI behind it (i.e. > {{hdfs balancer}}) exits with 0. This is probably caused by following piece > of code found in {{hadoop-daemon.sh}}: > {code:java} > sleep 3; > if ! ps -p $! > /dev/null ; then > exit 1 > fi > {code} > It seems the CLI command finishes so quickly in case of a balanced cluster > that the above {{ps}} does not find it. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org