Support Questions
Find answers, ask questions, and share your expertise
Announcements
Alert: Welcome to the Unified Cloudera Community. Former HCC members be sure to read and learn how to activate your account here.

Cloudera manager creating HDFS /tmp directory failed

Cloudera manager creating HDFS /tmp directory failed

New Contributor

I'm installing CDH5 on Centos 6.5 using Cloudera Manager 5.2. Every steps in "Cluster Installation" phase finished seamlessly. But in the last step of "Cluster Setup" phase, it keeps failing on "creating HDFS /tmp directory".

 
/tmp already exists as a directory created by the server.
I've looked at the script that is attempting to create the dir ( /usr/lib/cmf/service/hdfs/hdfs.sh ) and if the directory exists it should not even try to create the directory and keep going. Any ideas? Is there a way to change the hdfs /tmp dir?
 
I'm really curious why they would use /tmp in the first place, knowing EVERY server has it created already..
3 REPLIES 3
Highlighted

Re: Cloudera manager creating HDFS /tmp directory failed

Contributor

I believe this directory created is inside HDFS and the OS version of /tmp/ If you can get your cluster up and go to "hostname:50070" page it should show you the HDFS filesystem. TMP is created by default.

Re: Cloudera manager creating HDFS /tmp directory failed

New Contributor

My stderr log message is as follows:


+ /opt/cloudera/parcels/CDH-5.2.1-1.cdh5.2.1.p0.12/lib/hadoop-hdfs/bin/hdfs --config /var/run/cloudera-scm-agent/process/14-hdfs-NAMENODE-createtmp dfs -test -d /tmp
test: Call From centos65cm/192.168.142.129 to centos65cm:8020 failed on connection exception: java.net.ConnectException: Connection refused; For more details see:  http://wiki.apache.org/hadoop/ConnectionRefused
+ '[' 1 -eq 0 ']'
+ echo 'Directory /tmp does not currently exist.'
+ PARENT_ARG=
+ '[' 5 -ge 4 ']'
+ PARENT_ARG=-p
+ /opt/cloudera/parcels/CDH-5.2.1-1.cdh5.2.1.p0.12/lib/hadoop-hdfs/bin/hdfs --config /var/run/cloudera-scm-agent/process/14-hdfs-NAMENODE-createtmp dfs -mkdir -p /tmp
 
 
And the 8020 port is open since I can run netstat and lsof to check it:
 
[root@centos65cm ~]# netstat -tlpn |grep :8020
tcp        0      0 192.168.142.129:8020        0.0.0.0:*                   LISTEN      18392/java
[root@centos65cm ~]# lsof -i @centos65cm:8020
COMMAND   PID USER   FD   TYPE DEVICE SIZE/OFF NODE NAME
java    18392 hdfs  166u  IPv4 167822      0t0  TCP centos65cm:intu-ec-svcdisc (LISTEN)
 
 
The 8020 port is accessible since I can use telnet to connect to it: 
[root@centos65cm ~]# telnet centos65cm 8020
Trying 192.168.142.129...
Connected to centos65cm.
Escape character is '^]'.
 
 
I have four VM's. One is centos65cm, the other three centos65datanode[1-3]. I installed Cloudera Manager 5.2 on centos65cm, which is also the master node. All my IPs are configured as static. The /etc/hosts file look like this:
127.0.0.1   localhost localhost
192.168.142.129 centos65cm centos65cm
192.168.142.131 centos65datanode3 centos65datanode3
192.168.142.133 centos65datanode1 centos65datanode1
192.168.142.135 centos65datanode2 centos65datanode2
 
My four VM's are under NAT network adapter. I don't have any local domain configured.
 

But hdfs creating /tmp still failed. Any idea?

Re: Cloudera manager creating HDFS /tmp directory failed

Explorer

I had the same problem. When looking through the logs (start with agent log to see what script is being run) and then looking through the logs in the directory that the dfs create tmp script is in, I see that the directory is successfully created. However I suspect a timing issue between the agent and manager causes a command timeout. 

 

@CDH guys how do I file a PR?

==

 

cdh5.5.1 on ubuntu 14.04 trusty, install path B

Don't have an account?
Coming from Hortonworks? Activate your account here