graceful_stop.rb

puts('[spark] remove marker file by ensuring to trigger graceful stop spark-etl jobs')
`hadoop fs -rm -r /tmp/datalake.marker.*`

while true
  app_id = `yarn application -list | grep datalake.Main | awk -F " " '{print $1}'`.strip
  status = `yarn application -list | grep datalake.Main | awk -F " " '{print $6}'`.strip
  puts("[spark] application status: #{status}")

  if status == "ACCEPTED"
    puts("[spark] killing #{app_id}")
    `yarn application -kill #{app_id}`
    break
  end

  if status != 'RUNNING'
    puts("[spark] application '#{app_id}' was stopped")
    break
  end

  puts("[spark] application '#{app_id}' waiting graceful shutdown...")
  sleep(20)
end

puts('[spark] remove checkpoint directory because of rolling out the new application version')
`hadoop fs -rm -r /tmp/spark-etl/`