[BIGTOP-436] flume-node stop seems to mistarget some other java process on lucid

• Type: Bug
• Status: Open
• Priority: Critical
• Resolution: Unresolved
• Affects Version/s: 0.3.0
• Component/s: debian
• Assignee: Patrick Taylor Ramsey
• Reporter: Roman Shaposhnik
• Created: 08/Mar/12 16:46
• Updated: 26/Mar/13 18:12

Description

It seems that the following bit of code from flume-node reliably kills our jenkins slave process on lucid:

  # FLUME-919 will put an end to such extreme violence
  FLUME_PID=`cat $FLUME_PID_FILE`
  if [ -n $FLUME_PID ]; then
    FLUME_PID_GROUP=$(ps -o pgrp -p ${FLUME_PID} h)

    if [ -n $FLUME_PID_GROUP ]; then
      kill -TERM -${FLUME_PID_GROUP} &>/dev/null
      sleep 5
      kill -KILL -${FLUME_PID_GROUP} &>/dev/null

      rm -f $LOCKFILE $FLUME_PID_FILE
    fi
  fi
  return 0

Here's how it happens:
http://bigtop01.cloudera.org:8080/view/Bigtop-trunk/job/Bigtop-trunk-packagetest-lucid/label=lucid-slave/1/console

We need to investigate and possibly fix this.