INSTALL-HADOOP

#!/usr/bin/env bash



# curl -sSL https://raw.githubusercontent.com/DemandCube/hadoop-single-node-cluster/master/INSTALL-HADOOP | bash -s -- -r



FORCE=false
DEBUG=false
REMOTE=false
BLUEPRINT="blueprint-raw-all.json"
YARN=false
while getopts ":b:e:fdry" Option
# Initial declaration.
# a, b, c, d, e, f, and g are the flags expected.
# The : after flag 'e' shows it will have an option passed with it.
do
  case $Option in
#	w ) CMD=$OPTARG; FILENAME="PIMSLogList.txt"; TARGET="logfiles"; ;;
    e ) email_contact=$OPTARG ;;
    y ) BLUEPRINT="blueprint-raw-minimal.json" ;;
    b ) BLUEPRINT=$OPTARG ;;
	f ) FORCE=true ;;
	d ) DEBUG=true ;;
    r ) REMOTE=true ;;
	* ) echo "Not recognized argument"; exit -1  ;;
    
  esac
done
shift $(($OPTIND - 1))

if [ -f HADOOP-SINGLE-NODE-CLUSTER ] ; then
    cat HADOOP-SINGLE-NODE-CLUSTER
else
    if [ "$REMOTE" == "true" ] ; then
        curl -s https://raw.githubusercontent.com/DemandCube/hadoop-single-node-cluster/master/HADOOP-SINGLE-NODE-CLUSTER
        echo ""
    fi
fi
printf "\n\n"

sleep 1

# Check memory and if it's low warn the user
ram=$(free -l | head -n 2 | tail -n 1 | awk '{print $2}')
if [ "$ram" == "" ] ; then
    echo "WARNING: Problem checking how much ram you have."
else
    # test if machine has 4 gigs
    if [ $ram -lt 4194304 ] ; then
        echo "WARNING NOT ENOUGH MEMORY"
        echo "You have $(free -lg | head -n 2 | tail -n 1 | awk '{print $2}') gig of RAM"
        echo ""
        echo "RECOMMENDED: Minimum of 4 Gigs of RAM or will probably fail"
        echo ""
        if [ "$FORCE" == "false" ] ; then
            echo "Use: argument -f to force"
            exit 1;
        fi
    fi
fi



if [ $UID -ne 0 ] ; then
    echo "WARNING NOT BEING RUN AS ROOT"
    echo ""
    echo "RECOMMENDED: Run only as root, only tested that way."
    echo ""
    if [ "$FORCE" == "false" ] ; then
        echo "Use: argument -f to force"
        exit 1;
    fi
fi


if [ -f /etc/redhat-release ] ; then
    echo "Using OS:"
    cat /etc/redhat-release
else
    echo "WARNING this is only tested on CENTOS this may not work"
    printf "\n\n"
    if [ "$FORCE" == "false" ] ; then
        echo "Use: argument -f to force"
        exit 1;
    fi
fi


# email_contact=$1 # nagios wants the sysadmin's email address
if [[ $email_contact == "" ]] ; then
    echo "Usage: create_cluster.sh [someone@example.com]"
    echo "Optionally supply a contact email for Nagios monitoring."

    email_contact='nobody@noop.com'
    echo "Using default email for nagios: ${email_contact}"
    echo "Add argument to custom install email"
fi


function check_http_status(){
    url=$1
    target_status=$2
    local retstatus=0
    actual_status=$(curl -o /dev/null -s -w %{http_code} $url)
    
    if [[ $target_status == $actual_status ]]; then
            #echo "Target status matched."
            retstatus=0
    else
            #echo "Target status $target_status not equal to $actual_status"
            retstatus=1
    fi
    return $retstatus
}

function check_ambari() {
    local ambari_status=0
    local ps_ran=1
    if [ -f /var/run/ambari-server/ambari-server.pid ] ; then
        ps -p `cat /var/run/ambari-server/ambari-server.pid` &> /dev/null
    else
        ps_ran=0
    fi

    if [ $? -ne 0 ] || [ $ps_ran -eq 0 ]  ; then
        # Count 1 is not found
        # Count 2 is the search and process found
        PROCESS_CHECK=$(ps aux | grep org.apache.ambari.server.controller.AmbariServer | wc -l)
        if [ $PROCESS_CHECK -lt 2 ] ; then
            # Ambari's not running
            ambari_status=1
        fi
    fi
    return $ambari_status
}


function wait_until_some_http_status () {
    local url=$1
    local target_status=$2
    local check_ambari_running=$3
    local s=0
    local spinstr='|/-\'
    local clean_spin=0
    local seconds_past=0


    while [[ $s != $target_status ]]
        do
            if [ "x$check_ambari_running" == "xtrue" ] ; then
                check_ambari
                if [ $? -ne 0 ] ; then
                    echo "Ambari stopped running ending the loop"
                    exit 1
                fi
            fi
            s=$(curl -o /dev/null -s -w %{http_code} $url)
            if [[ $s == "000" ]]
                then
                    if [ ${REMOTE:-false} == "false" ] ; then  
                        if [ $clean_spin -eq 1 ] ; then
                            printf "\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b"
                        fi
                        local temp=${spinstr#?}
                        printf "no response from server - [%c]  " "$spinstr"
                        local spinstr=$temp${spinstr%"$temp"}
                        clean_spin=1
                    else 
                        echo "<no response from server>"
                    fi
                else
                    echo "HTTP status: $s"
            fi
            sleep 2
            let "seconds_past=$seconds_past+2"
            
            if [ $seconds_past -gt 120 ] ; then
                echo "waiting for ambari more than 2 minutes"
                echo "exiting now"
                exit 1
            fi

        done
}

spinner()
{
    local thetime=$1
    local delay=0.75
    local spinstr='|/-\'
    for (( c=0 ; c<=$thetime ; c++ )) {
        local temp=${spinstr#?}
        printf " [%c]  " "$spinstr"
        local spinstr=$temp${spinstr%"$temp"}
        sleep $delay
        printf "\b\b\b\b\b\b"
    }
    printf "    \b\b\b\b"
}

echo "First install and start needed services"
sudo yum install -y ntp ntpdate ntp-doc  # in many installs, ex. RHEL EC2 AMI, we already have ntpd and ntpupdate installed
sudo chkconfig ntpd on
sudo service ntpd restart

sudo chkconfig iptables off
sudo /etc/init.d/iptables stop
sudo chkconfig ip6tables off
sudo /etc/init.d/ip6tables stop

sudo yum update openssl.x86_64 -y

echo "The version of openssl.x86_64:"
yum list installed openssl.x86_64 | tail -n 1 | awk '{print $2}'

echo "openssl:"
echo "Good Version: 1.0.1e-16.el6_5.14 - (Known to work)"
echo "Bad Version: 1.0.1e-15.el6 - (Known to cause ambari agents to not register)"


# ---> Package openssl.x86_64 0:1.0.1e-15.el6 will be updated
# ---> Package openssl.x86_64 0:1.0.1e-16.el6_5.14 will be an update

echo "To get ambari with yum we need the Hortonworks repo"

# Old code using wget - replaced with the curl equivalent
# sudo yum install -y wget
# sudo wget -c -P /etc/yum.repos.d/ http://public-repo-1.hortonworks.com/ambari/centos6/1.x/updates/1.6.0/ambari.repo

curl --create-dirs -C - -o /etc/yum.repos.d/ambari.repo http://public-repo-1.hortonworks.com/ambari/centos6/1.x/updates/1.6.0/ambari.repo
sudo yum -y install ambari-server ambari-agent

# hostname -f
# hostname
# uname -n
# /etc/sysconfig/network
# /etc/hosts
# /etc/hostname

my_fqdn=$(python -c 'from socket import getfqdn; print getfqdn()')
echo "FQDN:$my_fqdn"
# Internally Ambari uses the following to get it's hostname
# >>> import socket
# >>> socket.getfqdn()
# 'localhost.localdomain'

ping -c 1 $my_fqdn

if [ ! $? -eq 0 ] ; then
my_fqdn=$(hostname)

# get the main ip address
# this gets the ip address that would be used to route to ip 8.8.8.8
my_ip=`ip route get 8.8.8.8 | awk 'NR==1 {print $NF}'`
my_short_fqdn=$my_fqdn
my_long_fqdn="${my_short_fqdn}.sandbox.neverwinterdp.com"

echo "FQDN:$my_fqdn"
echo "127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4" > /etc/hosts
echo "::1         localhost localhost.localdomain localhost6 localhost6.localdomain6" >> /etc/hosts
echo "$my_ip      ${my_long_fqdn} $my_fqdn" >> /etc/hosts
my_fqdn=$my_long_fqdn
fi
echo "FQDN:$my_fqdn"


sudo ambari-agent restart
sudo ambari-server setup -v -s
sudo ambari-server restart

TMP_DIR=/tmp

echo "Replace the dummy email address in the blueprint JSON file with the specified email address"

if [ "$REMOTE" == "true" ] ; then
    curl -s https://raw.githubusercontent.com/DemandCube/hadoop-single-node-cluster/master/$BLUEPRINT | sed s/NAGIOS_CONTACT_GOES_HERE/$email_contact/ > $TMP_DIR/blueprint.json
else
    sed s/NAGIOS_CONTACT_GOES_HERE/$email_contact/ $BLUEPRINT > $TMP_DIR/blueprint.json    
fi


echo "Trying http://localhost:8080/api/v1/blueprints to confirm Ambari server is up..."
wait_until_some_http_status "http://admin:admin@localhost:8080/api/v1/blueprints" "200"
sleep 2 # wait a few moments longer just to let the server settle down

echo "Add the blueprint.json blueprint file to our Ambari server's available blueprints"
curl -v -X POST -d @$TMP_DIR/blueprint.json http://admin:admin@localhost:8080/api/v1/blueprints/bp-all-services --header "Content-Type:application/json" --header "X-Requested-By:mycompany"

echo "Trying http://localhost:8080/api/v1/clusters to confirm Ambari server is still up..."
wait_until_some_http_status "http://admin:admin@localhost:8080/api/v1/clusters" "200" "true"

echo "Replacing the dummy hostname in the cluster creation JSON file with this host's fully qualified domain name"
# now set above
# my_fqdn=$(hostname -f)

if [ "$REMOTE" == "true" ] ; then
    curl -s https://raw.githubusercontent.com/DemandCube/hadoop-single-node-cluster/master/cluster-creation-raw.json | sed s/FQDN_GOES_HERE/$my_fqdn/ > $TMP_DIR/cluster-creation.json
else
    sed s/FQDN_GOES_HERE/$my_fqdn/ cluster-creation-raw.json > $TMP_DIR/cluster-creation.json
fi

echo ""
echo "Pausing for 30 seconds to let Ambari server settle down"

if [ "$REMOTE" == "true" ] ; then
    sleep 30
else
    spinner 40
fi


#if cluster already exists delete it

# Code by Anthony doesn't work
# if check_http_status  '-H "X-Requested-By: ambari" -u admin:admin -i  http://localhost:8080/api/v1/clusters/cl1' 200; then
#     echo "Cluster [cl1] already exists - deleting it"
#     curl -H "X-Requested-By: ambari" -u admin:admin -i -X DELETE http://localhost:8080/api/v1/clusters/cl1
# fi

check_http_status  ' -u admin:admin -i  http://localhost:8080/api/v1/clusters/cl1' 200

if [ $? -eq 0 ] ; then
    # curl -o /dev/null -s -w %{http_code} 
    echo "Cluster [cl1] already exists - deleting it"
    curl -H "X-Requested-By: ambari" -u admin:admin -i -X DELETE http://localhost:8080/api/v1/clusters/cl1
fi

echo ""
echo "Printing to see if the blueprint exists"
echo ""

curl http://admin:admin@localhost:8080/api/v1/blueprints/bp-all-services

echo ""
echo "Printing to see what register agent hosts exist"
echo ""

curl -uadmin:admin http://localhost:8080/api/v1/hosts


TEST_HOST=`curl -s -uadmin:admin http://localhost:8080/api/v1/hosts | grep $my_fqdn | grep host_name | awk '{print $3}' | tr -d [\"] | wc -l`

if [ $TEST_HOST -ne 1 ] ; then
    echo "WARNING: Ambari agent host isn't registered please investigate"
    echo "In the pass have found it to be an ssl error"
    echo "curl -uadmin:admin http://localhost:8080/api/v1/hosts"
fi


echo "Now cause a cluster to be created with our loaded blueprint"
curl -v -X POST -d @$TMP_DIR/cluster-creation.json http://admin:admin@localhost:8080/api/v1/clusters/cl1 --header "Content-Type:application/json" --header "X-Requested-By:mycompany"
echo ""
echo ""
echo "Cluster creation should now be in process. This script will now monitor it until completion. You can also point your browser to localhost:8080 and log in as admin:admin to monitor it yourself."
echo ""
# This gets COMPLETED when done
# curl -s -uadmin:admin http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=Requests/request_status | grep request_status | awk '{print $3}' | tr -d [\"]


CHECK_STATUS=true
LOOP_SECONDS=0

AMBARI_RESTARTS=0

while [[ "$CHECK_STATUS" == "true" ]] ; do
    
    CMD_CLUSTER_STATUS='curl -uadmin:admin -o /dev/null -s -w %{http_code} http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=Requests/request_status'
    CMD_CLUSTER_RESULTS='curl -s -uadmin:admin http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=Requests/request_status'
    CMD_CLUSTER_RESULTS_SECTION='curl -s -uadmin:admin http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=Requests'
    CMD_CLUSTER_TASK_SECTION='curl -s -uadmin:admin http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=tasks/Tasks/*'
        
    CLUSTER_STATUS=`$CMD_CLUSTER_STATUS`
    CLUSTER_RESULTS=`$CMD_CLUSTER_RESULTS`
    CLUSTER_RESULTS_STATUS=`echo "$CLUSTER_RESULTS" | grep request_status | awk '{print $3}' | tr -d [\"]`
    
    CLUSTER_TASK_RESULTS=`$CMD_CLUSTER_TASK_SECTION`
    CLUSTER_RESULTS_SECTION=`$CMD_CLUSTER_RESULTS_SECTION`
    
    CLUSTER_RESULTS_COMPLETED=`echo "$CLUSTER_RESULTS_SECTION" | grep completed_task_count | awk '{print $3}' | tr -d [\",]`
    CLUSTER_RESULTS_TASKS=`echo "$CLUSTER_RESULTS_SECTION" | grep \"task_count | awk '{print $3}' | tr -d [\",]`
    # echo "$CLUSTER_STATUS"
    # echo "$CLUSTER_RESULTS_STATUS"
    
    # Test if ambari server dies
    # /var/run/ambari-server/ambari-server.pid
    
    echo "Total Tasks:$CLUSTER_RESULTS_TASKS Tasks Completed:$CLUSTER_RESULTS_COMPLETED"
    
    ps -p `cat /var/run/ambari-server/ambari-server.pid` > /dev/null
    
    if [ $? -ne 0 ] ; then
        echo "Problem finding ambari running with pid file."
        if [ -f /var/run/ambari-server/ambari-server.pid ] ; then
            echo "Pid file is missing."
        else
            echo "found pid file but process is missing"
        fi
        # Count 1 is not found
        # Count 2 is the search and process found
        PROCESS_CHECK=$(ps aux | grep org.apache.ambari.server.controller.AmbariServer | wc -l)
        if [ $PROCESS_CHECK -lt 2 ] ; then
            echo "Ambari's not running restarting"
            sleep 10
            /etc/init.d/ambari-server start
            AMBARI_RESTARTS=$AMBARI_RESTARTS+1
            if [ $AMBARI_RESTARTS -ge 2 ] ; then
                echo "restarted ambari too many time check manually"
                exit 1
            fi
        fi
    fi
    
    
    # if pid missing or check with checking for process
    # /usr/jdk64/jdk1.7.0_45/bin/java -server -XX:NewRatio=3 -XX:+UseConcMarkSweepGC -XX:-UseGCOverheadLimit -XX:CMSInitiatingOccupancyFraction=60 -Xms512m -Xmx2048m -Djava.security.auth.login.config=/etc/ambari-server/conf/krb5JAASLogin.conf -Djava.security.krb5.conf=/etc/krb5.conf -Djavax.security.auth.useSubjectCredsOnly=false -cp /etc/ambari-server/conf:/usr/lib/ambari-server/*:/usr/lib64/qt-3.3/bin:/usr/local/sbin:/usr/local/bin:/sbin:/bin:/usr/sbin:/usr/bin:/root/bin:/usr/lib/ambari-server/*:/usr/share/java/mysql-connector-java.jar:/usr/share/java/mysql-connector-java-5.1.17.jar org.apache.ambari.server.controller.AmbariServer
    
    
if [[ "$CLUSTER_STATUS" == "000" ]] ; then
        if [ ${REMOTE:-false} == "false" ] ; then  
            if [ $clean_spin -eq 1 ] ; then
                printf "\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b"
            fi
            local temp=${spinstr#?}
            printf "no response from server - [%c]  " "$spinstr"
            local spinstr=$temp${spinstr%"$temp"}
            clean_spin=1
        else 
            echo "<no response from server, continue to poll> (seconds: $LOOP_SECONDS)"
        fi
elif [ "$CLUSTER_STATUS" == "200" ] ; then
    # Success and test for CLUSTER_RESULTS_STATUS (IN_PROGRESS,FAILED,COMPLETED)
    echo "Successful call (seconds: $LOOP_SECONDS) - current status:$CLUSTER_RESULTS_STATUS"
    
    if [ "$CLUSTER_RESULTS_STATUS" == "COMPLETED" ] || [ "$CLUSTER_RESULTS_STATUS" == "FAILED" ] ; then
        if [ "$CLUSTER_RESULTS_STATUS" == "COMPLETED" ] ; then
            echo "Done - Point browser to localhost:8080 and log in as admin:admin to use Ambari - Enjoy your cluster - NeverwinterDP Team - http://www.neverwinterdp.com"
            exit 0
        fi
        if [ "$CLUSTER_RESULTS_STATUS" == "FAILED" ] ; then
            echo "$CLUSTER_TASK_RESULTS"
            echo "It failed you'll need to investigate. Point browser to localhost:8080 and log in as admin:admin."
            exit 1
        fi
    fi
    
elif [ "$CLUSTER_STATUS" == "403" ] ; then
    # Check error message
    MSG="\nGot a 403 response so is probably a authentication problem.\n\n"
    printf "$MSG"
    echo "$CLUSTER_RESULTS" | grep message | awk '{print $3}' | tr -d [\"]
    exit 1
elif [ "$CLUSTER_STATUS" == "404" ] ; then
    MSG="\nGot a 404 response so is probably a bad query call.\n\n"
    printf "$MSG"
    echo "$CLUSTER_RESULTS" | grep message | awk '{print $3}' | tr -d [\"]
    exit 1
elif [ "${CLUSTER_STATUS:0:1}" == "2" ] ; then
    MSG="\nGot a 2XX response so should be good.\n\n"
    printf "$MSG"
    
    echo "$CLUSTER_RESULTS"
    
    printf "$MSG"
    echo "Ending Loop"
    exit 1
    
elif [ "${CLUSTER_STATUS:0:1}" == "4" ] ; then
    MSG="\nGot a 4XX response so there is probably a problem.\n\n"
    printf "$MSG"
    
    echo "$CLUSTER_RESULTS"
    
    printf "$MSG"
    exit 1
else
    MSG="\nGot a unknown response so there is probably a problem.\n\n"
    printf "$MSG"
    echo "CLUSTER_RESULTS: ${CLUSTER_RESULTS}"
    echo "HTTP_STATUS: ${CLUSTER_STATUS}"
    echo "JSON_RESPONSE_STATUS: ${CLUSTER_RESULTS_STATUS}"
fi

let "LOOP_SECONDS=$LOOP_SECONDS+1"

sleep 1

done


# {
#   "href" : "http://localhost:8080/api/v1/clusters/cl1/requests/1",
#   "Requests" : {
#     "aborted_task_count" : 0,
#     "cluster_name" : "cl1",
#     "completed_task_count" : 0,
#     "create_time" : 1402883226676,
#     "end_time" : -1,
#     "failed_task_count" : 0,
#     "id" : 1,
#     "inputs" : null,
#     "progress_percent" : 4.757575757575758,
#     "queued_task_count" : 31,
#     "request_context" : "Install and start all services",
#     "request_schedule" : null,
#     "request_status" : "IN_PROGRESS",
#     "resource_filters" : [ ],
#     "start_time" : 1402883226951,
#     "task_count" : 66,
#     "timed_out_task_count" : 0,
#     "type" : "INTERNAL_REQUEST"
#   }
# }

#  echo "$CLUSTER_RESULTS"
# {
#   "status" : 404,
#   "message" : "Request resource doesn't exist."
# }

# curl: (6) Couldn't resolve host '%http_code'
# {
#   "status": 403,
#   "message": "Full authentication is required to access this resource"
# }

# [root@test2 ~]# curl -s -uadmin:admin http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=Requests/request_status
# {
#   "href" : "http://localhost:8080/api/v1/clusters/cl1/requests/1",
#   "Requests" : {
#     "cluster_name" : "cl1",
#     "id" : 1,
#     "request_status" : "IN_PROGRESS"
#   }
# }

# [root@ip-10-233-133-34 ~]# curl -s -uadmin:admin http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=Requests/request_status
# {
#   "status" : 404,
#   "message" : "Request resource doesn't exist."
# }

# [vagrant@sparkngin1 ~]$ curl -s -uadmin:admin http://localhost:8080/api/v1/clusters/cl1/requests/1?fields=Requests/request_status
# {
#   "href" : "http://localhost:8080/api/v1/clusters/cl1/requests/1",
#   "Requests" : {
#     "cluster_name" : "cl1",
#     "id" : 1,
#     "request_status" : "FAILED"
#   }
# }







# [root@ip-10-36-3-46 ~]# curl -v -X POST -d @$TMP_DIR/cluster-creation.json http://admin:admin@localhost:8080/api/v1/clusters/cl1 --header "Content-Type:application/json" --header "X-Requested-By:mycompany"
# * About to connect() to localhost port 8080 (#0)
# *   Trying 127.0.0.1... connected
# * Connected to localhost (127.0.0.1) port 8080 (#0)
# * Server auth using Basic with user 'admin'
# > POST /api/v1/clusters/cl1 HTTP/1.1
# > Authorization: Basic YWRtaW46YWRtaW4=
# > User-Agent: curl/7.19.7 (x86_64-redhat-linux-gnu) libcurl/7.19.7 NSS/3.14.0.0 zlib/1.2.3 libidn/1.18 libssh2/1.4.2
# > Host: localhost:8080
# > Accept: */*
# > Content-Type:application/json
# > X-Requested-By:mycompany
# > Content-Length: 251
# > 
# < HTTP/1.1 400 Bad Request
# < Set-Cookie: AMBARISESSIONID=1loev1y9bqi551lcirxdyqs21q;Path=/
# < Expires: Thu, 01 Jan 1970 00:00:00 GMT
# < Content-Type: text/plain
# < Content-Length: 178
# < Server: Jetty(7.6.7.v20120910)
# < 
# {
#   "status" : 400,
#   "message" : "Attempted to add unknown hosts to a cluster.  These hosts have not been registered with the server: ip-10-36-3-46.us-west-2.compute.internal"
# * Connection #0 to host localhost left intact
# * Closing connection #0
# }


# [root@ip-10-36-19-68 ~]# curl -uadmin:admin http://localhost:8080/api/v1/hosts
# {
#   "href" : "http://localhost:8080/api/v1/hosts",
#   "items" : [
#     {
#       "href" : "http://localhost:8080/api/v1/hosts/ip-10-36-19-68.us-west-2.compute.internal",
#       "Hosts" : {
#         "host_name" : "ip-10-36-19-68.us-west-2.compute.internal"
#       }
#     }
#   ]
# }