Customizing cluster and application configuration with earlier AMI versions of Amazon EMR
Amazon EMR release version 4.0.0 introduced a simplified method of configuring applications
using configuration classifications. For more information, see Configure applications. When using an AMI
version, you configure applications using bootstrap actions along with arguments that
you pass. For example, the configure-hadoop and
configure-daemons bootstrap actions set Hadoop and YARN–specific
environment properties like --namenode-heap-size. In more recent versions,
these are configured using the hadoop-env and yarn-env
configuration classifications. For bootstrap actions that perform common configurations,
see the emr-bootstrap-actions repository on Github
The following tables map bootstrap actions to configuration classifications in more recent Amazon EMR release versions.
Hadoop | ||
|---|---|---|
| Affected application file name | AMI version bootstrap action | Configuration classification |
core-site.xml |
configure-hadoop -c
|
core-site |
log4j.properties |
configure-hadoop -l |
hadoop-log4j |
hdfs-site.xml |
configure-hadoop -s |
hdfs-site
|
| n/a | n/a | hdfs-encryption-zones |
mapred-site.xml
|
configure-hadoop -m |
mapred-site |
yarn-site.xml
|
configure-hadoop -y
|
yarn-site |
httpfs-site.xml |
configure-hadoop -t |
httpfs-site |
capacity-scheduler.xml
|
configure-hadoop -z
|
capacity-scheduler |
yarn-env.sh |
configure-daemons --resourcemanager-opts |
yarn-env |
Hive | ||
|---|---|---|
| Affected application file name | AMI version bootstrap action | Configuration classification |
hive-env.sh |
n/a | hive-env |
hive-site.xml |
hive-script --install-hive-site
${MY_HIVE_SITE_FILE} |
hive-site |
hive-exec-log4j.properties |
n/a | hive-exec-log4j |
hive-log4j.properties |
n/a | hive-log4j |
EMRFS | ||
|---|---|---|
| Affected application file name | AMI version bootstrap action | Configuration classification |
emrfs-site.xml |
configure-hadoop -e |
emrfs-site |
| n/a | s3get -s s3://custom-provider.jar -d
/usr/share/aws/emr/auxlib/ |
emrfs-site (with new setting
fs.s3.cse.encryptionMaterialsProvider.uri) |
For a list of all classifications, see Configure applications.
Application environment variables
When using an AMI version, a hadoop-user-env.sh
script is used along with the configure-daemons bootstrap action to
configure the Hadoop environment. The script includes the following actions:
#!/bin/bash export HADOOP_USER_CLASSPATH_FIRST=true; echo "HADOOP_CLASSPATH=/path/to/my.jar" >> /home/hadoop/conf/hadoop-user-env.sh
In Amazon EMR release 4.x, you do the same using the hadoop-env
configuration classification, as shown in the following example:
[ { "Classification":"hadoop-env", "Properties":{ }, "Configurations":[ { "Classification":"export", "Properties":{ "HADOOP_USER_CLASSPATH_FIRST":"true", "HADOOP_CLASSPATH":"/path/to/my.jar" } } ] } ]
As another example, using configure-daemons and passing --namenode-heap-size=2048 and
--namenode-opts=-XX:GCTimeRatio=19 is equivalent to the following configuration classifications.
[ { "Classification":"hadoop-env", "Properties":{ }, "Configurations":[ { "Classification":"export", "Properties":{ "HADOOP_DATANODE_HEAPSIZE": "2048", "HADOOP_NAMENODE_OPTS": "-XX:GCTimeRatio=19" } } ] } ]
Other application environment variables are no longer defined in
/home/hadoop/.bashrc. Instead, they are primarily set in
/etc/default files per component or application, such as
/etc/default/hadoop. Wrapper scripts in
/usr/bin/ installed by application RPMs may also set additional
environment variables before involving the actual bin script.
Service ports
When using an AMI version, some services use custom ports.
Changes in port settings | ||
|---|---|---|
| Setting | AMI version 3.x | Open-source default |
| fs.default.name | hdfs://emrDeterminedIP:9000 | default (hdfs://emrDeterminedIP:8020)
|
| dfs.datanode.address | 0.0.0.0:9200 | default (0.0.0.0:50010) |
| dfs.datanode.http.address | 0.0.0.0:9102 | default (0.0.0.0:50075) |
| dfs.datanode.https.address | 0.0.0.0:9402 | default (0.0.0.0:50475) |
| dfs.datanode.ipc.address | 0.0.0.0:9201 | default (0.0.0.0:50020) |
| dfs.http.address | 0.0.0.0:9101 | default (0.0.0.0:50070) |
| dfs.https.address | 0.0.0.0:9202 | default (0.0.0.0:50470) |
| dfs.secondary.http.address | 0.0.0.0:9104 | default (0.0.0.0:50090) |
| yarn.nodemanager.address | 0.0.0.0:9103 | default (${yarn.nodemanager.hostname}:0) |
| yarn.nodemanager.localizer.address | 0.0.0.0:9033 | default (${yarn.nodemanager.hostname}:8040) |
| yarn.nodemanager.webapp.address | 0.0.0.0:9035 | default (${yarn.nodemanager.hostname}:8042) |
| yarn.resourcemanager.address | emrDeterminedIP:9022 |
default (${yarn.resourcemanager.hostname}:8032) |
| yarn.resourcemanager.admin.address | emrDeterminedIP:9025 |
default (${yarn.resourcemanager.hostname}:8033) |
| yarn.resourcemanager.resource-tracker.address | emrDeterminedIP:9023 |
default (${yarn.resourcemanager.hostname}:8031) |
| yarn.resourcemanager.scheduler.address | emrDeterminedIP:9024 |
default (${yarn.resourcemanager.hostname}:8030) |
| yarn.resourcemanager.webapp.address | 0.0.0.0:9026 | default (${yarn.resourcemanager.hostname}:8088) |
| yarn.web-proxy.address | emrDeterminedIP:9046 |
default (no-value) |
| yarn.resourcemanager.hostname | 0.0.0.0 (default) | emrDeterminedIP |
The emrDeterminedIP is an IP address that is generated by
Amazon EMR.
Users
When using an AMI version, the user hadoop runs
all processes and owns all files. In Amazon EMR release version 4.0.0 and later, users
exist at the application and component level.
Installation sequence, installed artifacts, and log file locations
When using an AMI version, application artifacts and their configuration
directories are installed in the
/home/hadoop/
directory. For example, if you installed Hive, the directory would be
application/home/hadoop/hive. In Amazon EMR release 4.0.0 and later,
application artifacts are installed in the
/usr/lib/
directory. When using an AMI version, log files are found in various places. The
table below lists locations.application
Changes in log locations on Amazon S3 | |
|---|---|
| Daemon or application | Directory location |
| instance-state | node/instance-id/instance-state/ |
| hadoop-hdfs-namenode | daemons/instance-id/hadoop-hadoop-namenode.log |
| hadoop-hdfs-datanode | daemons/instance-id/hadoop-hadoop-datanode.log |
| hadoop-yarn (ResourceManager) | daemons/instance-id/yarn-hadoop-resourcemanager |
| hadoop-yarn (Proxy Server) | daemons/instance-id/yarn-hadoop-proxyserver |
| mapred-historyserver | daemons/instance-id/ |
| httpfs | daemons/instance-id/httpfs.log |
| hive-server | node/instance-id/hive-server/hive-server.log |
| hive-metastore | node/instance-id/apps/hive.log |
| Hive CLI | node/instance-id/apps/hive.log |
| YARN applications user logs and container logs | task-attempts/ |
| Mahout | N/A |
| Pig | N/A |
| spark-historyserver | N/A |
| mapreduce job history files | jobs/ |
Command runner
When using an AMI version, many scripts or programs, like
/home/hadoop/contrib/streaming/hadoop-streaming.jar, are
not placed on the shell login path environment, so you need to specify the full path
when you use a jar file such as command-runner.jar or script-runner.jar to execute
the scripts. The command-runner.jar is located on the AMI so
there is no need to know a full URI as was the case with
script-runner.jar.
Replication factor
The replication factor lets you configure when to start a Hadoop JVM. You can start a new
Hadoop JVM for every task, which provides better task isolation, or you can share
JVMs between tasks, providing lower framework overhead. If you are processing many
small files, it makes sense to reuse the JVM many times to amortize the cost of
start-up. However, if each task takes a long time or processes a large amount of
data, then you might choose to not reuse the JVM to ensure that all memory is freed
for subsequent tasks. When using an AMI version, you can customize the replication
factor using the configure-hadoop bootstrap action to set the
mapred.job.reuse.jvm.num.tasks property.
The following example demonstrates setting the JVM reuse factor for infinite JVM reuse.
Linux line continuation characters (\) are included for readability. They can be removed or used in Linux commands. For Windows, remove them or replace with a caret (^).
aws emr create-cluster --name "Test cluster" --ami-version3.11.0\ --applications Name=HueName=HiveName=Pig\ --use-default-roles --ec2-attributes KeyName=myKey\ --instance-groups InstanceGroupType=MASTER,InstanceCount=1,InstanceType=m3.xlarge\ InstanceGroupType=CORE,InstanceCount=2,InstanceType=m3.xlarge\ --bootstrap-actions Path=s3://elasticmapreduce/bootstrap-actions/configure-hadoop,\ Name="Configuring infinite JVM reuse",Args=["-m","mapred.job.reuse.jvm.num.tasks=-1"]
