Ambari can't start namenode with a connection exception - hadoop

Question

I have been strapped in this for a few days,I tried a lot of methods on the Internet but it didn't work,so I register on stackoverflow and write my first question.
My environment is HDP2.4 and Ubuntu14.04 LTS
the log information is like this:
safemode: Call From master/9.119.131.105 to master:8020 failed on connection exception: java.net.ConnectException: Connection refused; For more details see: http://wiki.apache.org/hadoop/ConnectionRefused
2016-04-21 15:59:57,796 - Retrying after 10 seconds. Reason: Execution of '/usr/hdp/current/hadoop-hdfs-namenode/bin/hdfs dfsadmin -fs hdfs://master:8020 -safemode get | grep 'Safe mode is OFF'' returned 1. safemode: Call From master/9.119.131.105 to master:8020 failed on connection exception: java.net.ConnectException: Connection refused; For more details see: http://wiki.apache.org/hadoop/ConnectionRefused
(1).At first I tried to avoid the safemode,but it isn't the key of this issue.
(2).Then I started to work on the 8020 port of master,master is the namenode of my cluster and I am sure that the hosts file is correct:
9.119.131.105 master
(3).ssh function is good,the four nodes in the cluster can load each other without passcodes
(4).I'm sure the fire wall is off
# /etc/init.d/iptables stop
# ufw status
Status: inactive
(5).I also tried to open the 8020 port manually:
# iptables -A INPUT -p tcp --dport 8020 -j ACCEPT
it still didn't work.....
(6).I tried :
# telnet master 8020
but:
telnet:Unable to connect to remote host: Connect refused
(7).I find that not only the 8020 port can't work,but also 50070 can't work
(8).I'm sure the parameters in the configure files is 8020:
fs.defaultFS
hdfs://master:8020
dfs.namenode.rpc-address
master:8020
I'm expecting somebody can help me,I will be very appreciate about thant.
Thank you!