In the last post, I discussed the steps to install a 3 node hadoop cluster by using Cloudera Manager. In the next few posts, I am going to discuss some technologies that are frequently used, such as Hive, Sqoop, Impala and Spark.
There are a few things that need to be configured after the CDH Installation.
1. Configure NTPD. Start up ntpd process on every host. Otherwise, Clouder Manager could display a healthcheck failure: The host’s NTP service did not respond to a request for the clock offset.
# service ntpd status
# service ntpd start
# chkconfig ntpd on
# chkconfig –list ntpd
# ntpdc -np
2. Configure Replication Factor. As my little cluster has only 2 Data nodes, I need to reduce the replication factor from the default value of 3 to 2 to avoid the annoying blocks under-replicated type of error. First run the following command to change the replication factor to 2.
hadoop fs -setrep -R 2 /
Then goto HDFS Configuration, change Replication Factor to 2.
3. Change message logging level from INFO to WARN. I can not believe how many INFO messages are logged and there are no way I can see a message for more than 3 seconds before it is quickly refreshed away by a flood of INFO messages. In my opinion, majority of the INFO messages are useless and should not be logged in the first place. It seems more like DEBUG messages to me. So before my little cluster goes crazy in logging tons of useless messages, I need to quickly change logging level from INFO to WARNING. Another painful thing is that there are many log files from various Hadoop components, and are located at many different locations. I feel like I am siting in a space shuttle cockpit and need to turn off many switches not in a central location.
I could find out the logfile configuration file, and fix the parameters one by one. But it would take some time and too painful. The easiest way I found out is to use Cloudera Manager to make the change. Bascially, type in logging level as the search term. It will pop up a long list of components with the logging level and change them one by one. You will not believe how many logging level parameters are in the system. After the change, it’s recommended to restart the cluster as certain parameters are stale.
4. Configure Hue’s superuser and password. From Cloudera Manager screen, click Hue to start the Hue screen. The weird part about the Hue is that there is no pre-set superuser for the administration. Whoever logon to the Hue first will become the superuser of Hue. I don’t understand why Hue just takes whatever user and password Cloudera Manager uses. Anyway, to make my life easier, I just use the same login user and password for Cloudera Manager, admin.
5. Add new user.
By default hdfs user is the superuser for HDFS, not the root user. So before doing any work on Hadoop, it is a good idea to create a separte OS user instead of using hdfs user to execute Hadoop commands. Run the following commands on EVERY Host in the cluster.
a. Logon as root user.
b. Create bigdata group.
# groupadd bigdata
# grep bigdata /etc/group
c. Add the new user, wzhou.
# useradd -G bigdata -m wzhou
If the user exist before the bigdata created, do the following
# usermod -a -G bigdata wzhou
d. Change password
# passwd wzhou
e. Verify the user.
# id wzhou
f. Create the user home directory on HDFS.
# sudo -u hdfs hdfs dfs -mkdir /user/wzhou
# sudo -u hdfs hdfs dfs -ls /user
[root@vmhost1 ~]# sudo -u hdfs hdfs dfs -ls /user Found 8 items drwxrwxrwx - mapred hadoop 0 2015-09-15 05:40 /user/history drwxrwxr-t - hive hive 0 2015-09-15 05:44 /user/hive drwxrwxr-x - hue hue 0 2015-09-15 10:12 /user/hue drwxrwxr-x - impala impala 0 2015-09-15 05:46 /user/impala drwxrwxr-x - oozie oozie 0 2015-09-15 05:47 /user/oozie drwxr-x--x - spark spark 0 2015-09-15 05:41 /user/spark drwxrwxr-x - sqoop2 sqoop 0 2015-09-15 05:42 /user/sqoop2 drwxr-xr-x - hdfs supergroup 0 2015-09-20 11:23 /user/wzhou
g. Change the ownership of the directory.
# sudo -u hdfs hdfs dfs -chown wzhou:bigdata /user/wzhou
# hdfs dfs -ls /user
[root@vmhost1 ~]# sudo -u hdfs hdfs dfs -chown wzhou:bigdata /user/wzhou [root@vmhost1 ~]# sudo -u hdfs hdfs dfs -ls /user</strong> Found 8 items drwxrwxrwx - mapred hadoop 0 2015-09-15 05:40 /user/history drwxrwxr-t - hive hive 0 2015-09-15 05:44 /user/hive drwxrwxr-x - hue hue 0 2015-09-15 10:12 /user/hue drwxrwxr-x - impala impala 0 2015-09-15 05:46 /user/impala drwxrwxr-x - oozie oozie 0 2015-09-15 05:47 /user/oozie drwxr-x--x - spark spark 0 2015-09-15 05:41 /user/spark drwxrwxr-x - sqoop2 sqoop 0 2015-09-15 05:42 /user/sqoop2 drwxr-xr-x - wzhou bigdata 0 2015-09-20 11:23 /user/wzhou
h. Run a sample test.
Logon as wzhou user and verify whether the user can run sample MapReduce job from hadoop-mapreduce-examples.jar.
[wzhou@vmhost1 hadoop-mapreduce]$ hadoop jar /usr/lib/hadoop-mapreduce/hadoop-mapreduce-examples.jar pi 10 1000000 Number of Maps = 10 Samples per Map = 1000000 Wrote input for Map #0 Wrote input for Map #1 Wrote input for Map #2 Wrote input for Map #3 Wrote input for Map #4 Wrote input for Map #5 Wrote input for Map #6 Wrote input for Map #7 Wrote input for Map #8 Wrote input for Map #9 Starting Job 15/09/20 11:32:28 INFO client.RMProxy: Connecting to ResourceManager at vmhost1.local/192.168.56.71:8032 15/09/20 11:32:29 INFO input.FileInputFormat: Total input paths to process : 10 15/09/20 11:32:29 INFO mapreduce.JobSubmitter: number of splits:10 15/09/20 11:32:29 INFO mapreduce.JobSubmitter: Submitting tokens for job: job_1442764085933_0001 15/09/20 11:32:30 INFO impl.YarnClientImpl: Submitted application application_1442764085933_0001 15/09/20 11:32:30 INFO mapreduce.Job: The url to track the job: http://vmhost1.local:8088/proxy/application_1442764085933_0001/ 15/09/20 11:32:30 INFO mapreduce.Job: Running job: job_1442764085933_0001 15/09/20 11:32:44 INFO mapreduce.Job: Job job_1442764085933_0001 running in uber mode : false 15/09/20 11:32:44 INFO mapreduce.Job: map 0% reduce 0% 15/09/20 11:32:55 INFO mapreduce.Job: map 10% reduce 0% 15/09/20 11:33:03 INFO mapreduce.Job: map 20% reduce 0% 15/09/20 11:33:11 INFO mapreduce.Job: map 30% reduce 0% 15/09/20 11:33:18 INFO mapreduce.Job: map 40% reduce 0% 15/09/20 11:33:26 INFO mapreduce.Job: map 50% reduce 0% 15/09/20 11:33:34 INFO mapreduce.Job: map 60% reduce 0% 15/09/20 11:33:42 INFO mapreduce.Job: map 70% reduce 0% 15/09/20 11:33:50 INFO mapreduce.Job: map 80% reduce 0% 15/09/20 11:33:58 INFO mapreduce.Job: map 90% reduce 0% 15/09/20 11:34:06 INFO mapreduce.Job: map 100% reduce 0% 15/09/20 11:34:14 INFO mapreduce.Job: map 100% reduce 100% 15/09/20 11:34:14 INFO mapreduce.Job: Job job_1442764085933_0001 completed successfully 15/09/20 11:34:15 INFO mapreduce.Job: Counters: 49 File System Counters FILE: Number of bytes read=124 FILE: Number of bytes written=1258521 FILE: Number of read operations=0 FILE: Number of large read operations=0 FILE: Number of write operations=0 HDFS: Number of bytes read=2680 HDFS: Number of bytes written=215 HDFS: Number of read operations=43 HDFS: Number of large read operations=0 HDFS: Number of write operations=3 Job Counters Launched map tasks=10 Launched reduce tasks=1 Data-local map tasks=10 Total time spent by all maps in occupied slots (ms)=65668 Total time spent by all reduces in occupied slots (ms)=6387 Total time spent by all map tasks (ms)=65668 Total time spent by all reduce tasks (ms)=6387 Total vcore-seconds taken by all map tasks=65668 Total vcore-seconds taken by all reduce tasks=6387 Total megabyte-seconds taken by all map tasks=67244032 Total megabyte-seconds taken by all reduce tasks=6540288 Map-Reduce Framework Map input records=10 Map output records=20 Map output bytes=180 Map output materialized bytes=360 Input split bytes=1500 Combine input records=0 Combine output records=0 Reduce input groups=2 Reduce shuffle bytes=360 Reduce input records=20 Reduce output records=0 Spilled Records=40 Shuffled Maps =10 Failed Shuffles=0 Merged Map outputs=10 GC time elapsed (ms)=1026 CPU time spent (ms)=8090 Physical memory (bytes) snapshot=3877482496 Virtual memory (bytes) snapshot=17644212224 Total committed heap usage (bytes)=3034685440 Shuffle Errors BAD_ID=0 CONNECTION=0 IO_ERROR=0 WRONG_LENGTH=0 WRONG_MAP=0 WRONG_REDUCE=0 File Input Format Counters Bytes Read=1180 File Output Format Counters Bytes Written=97 Job Finished in 106.368 seconds Estimated value of Pi is 3.14158440000000000000
To restart all services in the cluster, you can just click Restart Action on the cluster from Cloudera Manager screen. However, if you want to start/stop a particular service, you might want to know the dependency of the services. Here are the order of starting/stopping sequence for all services on CDH 5.
1. Cloudera Management service
7. Key-Value Store Indexer
8. MapReduce or YARN
6. MapReduce or YARN
7. Key-Value Store Indexer
13. Cloudera Management Service
Ok, we are good here. In the next post, I am going to discuss load data to Hive.