#!/bin/sh

#yarn job status：NEW、NEW_SAVING、SUBMITTED、ACCEPTED、RUNNING、FINISHED、FAILED、KILLED

nnum=$(xx/yarn application -list -appStates NEW | wc -l)

nsnum=$(xx/yarn application -list -appStates NEW_SAVING | wc -l)

snum=$(xx/yarn application -list -appStates SUBMITTED | wc -l)

anum=$(xx/yarn application -list -appStates ACCEPTED | wc -l)

rnum=$(/xx/yarn application -list -appStates RUNNING | wc -l)

finum=$(xx/yarn application -list -appStates FINISHED | wc -l)

fanum=$(xxx/yarn application -list -appStates FAILED | wc -l)

knum=$(xxx/yarn application -list -appStates KILLED | wc -l)

let nnum=nnum-2

let nsnum=nsnum-2

let snum=snum-2

let anum=anum-2

let rnum=rnum-2

let finum=finum-2

let fanum=fanum-2

let knum=knum-2

vnum=($nnum $nsnum $snum $anum $rnum $finum $fanum $knum)

vkey=(NEW NEW_SAVING SUBMITTED ACCEPTED RUNNING FINISHED FAILED KILLED)

url="http://xxx/monitor/report/sendReport"

#yarn applicaiton -status application_1554987290796_77895

for (( i = 0 ; i < ${#vnum[@]} ; i++ ))

do

  for (( j = 0 ; j < ${#vkey[@]} ; j++ ))

  do

  if [ $i -eq $j ]

  then

  key="xx.yarn_job_state_"${vkey[$j]}"_count##xxx"

  value=${vnum[$i]}

  data="{\"key\":\"$key\",\"value\":$value}"

  echo $data

  cmd="curl -s -X POST -H 'Content-type':'application/json' -d '${data}' ${url}"

  sh -c "$cmd"

  fi

  done

done

#!/bin/sh

ip=xxx

port=8088

export HADOOP_HOME=/app/hadoop/bin

rmstate1=$($HADOOP_HOME/yarn rmadmin -getServiceState rm1)

rmstate2=$($HADOOP_HOME/yarn rmadmin -getServiceState rm2)

if [[ $rmstate1 == *active* ]]

then

   ip=xxx

fi

$HADOOP_HOME/yarn application -list -appStates RUNNING | awk '{print $1}' | grep -i "application" >appname.txt

curl http://$ip:$port/ws/v1/cluster/scheduler | jq . | sed 's/\"//g' > scheduler.json

sed -i '1,5d' scheduler.json

sed -n -e :a -e '1,33!{P;N;D;};N;ba' scheduler.json > nscheduler.json

cat nscheduler.json | grep -i "queueName" | awk -F ": " '{print $2}' | awk -F "," '{print $1}' > qname.txt

cat nscheduler.json | grep -i "memory" | awk -F ": " '{print $2}' | awk -F "," '{print $1}' > memory.txt

line=$(cat memory.txt | wc -l)

for((j=2;j<$line;j=j+6))

do

  sed -n '$jp' memory.txt >> maxmemory.txt

done

app=($(cat appname.txt | awk '{print $1}'))

for (( i = 0 ; i < ${#app[@]} ; i++ ))

do

    queue=$(curl http://$ip:$port/ws/v1/cluster/apps/${app[$i]} | jq . | sed 's/\"//g'  | grep -i "queue" | awk -F ":" '{print $2}')

    allocatedMB=$(curl http://$ip:$port/ws/v1/cluster/apps/${app[$i]} | jq . | sed 's/\"//g'  | grep -i "allocatedMB" | awk -F ":" '{print $2}')

    nparr=($(cat nscheduler.json | grep -i "numPendingApps" | awk -F ": " '{print $2}' | awk -F "," '{print $1}'))

    for((j=0;j<=${#nparr[@]};j++))

    do

       if [${nparr[$j]} -gt 10 ]

       then

          qname=$(sed -n '$jp' qname.txt)

          mem=$(sed -n '$jp' maxmemory.txt)

          if [ $qname -eq $queue ]

          then

                 rate=$(echo "sclae=2; $allocatedMB/$mem" | bc)

                 if [ $rate -gt 0.2 ]

                 then

                    $HADOOP_HOME/yarn application  -kill  ${app[$i]}

                 fi

          fi

        fi

    done

done

#查看任务的详细信息

#$HADOOP_HOME/yarn applicaiton -status ${app[$i]}

curl http://$ip:$port/ws/v1/cluster/apps/application_1554987290796_688678

#查看任务状态

curl http://$ip:$port/ws/v1/cluster/apps/application_1409421698529_0012/state

#kill任务

curl -v -X PUT -d '{"state": "KILLED"}''http://$ip:$port/ws/v1/cluster/apps/application_1409421698529_0012'

#队列资源信息

curl http://$ip:$port/ws/v1/cluster/scheduler

#查看指定队列的所有任务：

curl http://$ip:$port/ws/v1/cluster/apps?queue=default

#查看整个集群的指标

curl http://$ip:$port/ws/v1/cluster/metrics

#查看队列的状态

yarn queue -status default