This instruction can be applied both for single and multi-node deployment. It is based on the article: https://community.hortonworks.com/articles/60805/deploying-a-fresh-metron-cluster-using-ambari-serv.html and described how-to install HDP 2.5 with Metron.
Check it out if anything goes wrong or in troubleshooting section below. There is troubleshooting section in this article as well, make sure to check it first.
IMPORTANT: Some of the values specified in this article need to be substituted with your info, like IP addresses and hostnames. Such values are enclosed with brackers <like that>. Please make sure replace them with your info.
Centos 6
In order to ease install make sure that your main interface is named eth0. If it is not, you need to adjust some Ambari services configuration accordingly (e.g. ElasticSearch)
Single-node: At least 48 GB RAM, 8 cores and 400 GB HDD. Multi-node: At least 32 GB RAM, 4 cores and 200 GB HDD for smooth performance.
(optional) Disable PackageKit if it is installed, if it is not, just ignore this point:
sed -i 's/enabled=1/enabled=0/g' /etc/yum/pluginconf.d/refresh-packagekit.conf |
Increase limits for ElasticSearch and Storm on nodes where you will be installing them (if you don't know, increase it everywhere):
echo -e "elasticsearch - memlock unlimited\nstorm - nproc 257597" >> /etc/security/limits.conf |
Disable IPv6, leaving it enabled may force service to bind to IPv6 addresses only and thus resulting in inability to connect to it (source link):
sysctl -w net.ipv6.conf.all.disable_ipv6=1 sysctl -w net.ipv6.conf.default.disable_ipv6=1 echo -e "\n# Disable IPv6\nnet.ipv6.conf.all.disable_ipv6 = 1\nnet.ipv6.conf.default.disable_ipv6 = 1" >> /etc/sysctl.conf |
Disable Transparent Hugepage. Add "transparent_hugepage=never" to the end of the kernel line in grub.conf and reboot. (Ambari demands it, do we need to comply?):
transparent_hugepage=never |
After reboot check that changes were applied (make sure that word "never" is selected in square-brackets):
# cat /sys/kernel/mm/transparent_hugepage/enabled always madvise [never] |
On all nodes Install pre-requisites for Ambari:
yum install epel-release -y yum update -y yum install git wget curl rpm scp tar unzip bzip2 wget createrepo reposync yum-utils ntp python-pip -y |
On main node where your Ambari will be located install following:
pip install --upgrade pip pip install --upgrade setuptools |
Install Maven on main node and on Metron node install java 1.8 (if you don't know what it is - install it everywhere):
yum install java-1.8.0-openjdk java-1.8.0-openjdk-devel -y |
Set path to Java 8 if it does not exist:
export JAVA_HOME=$(readlink -f /usr/bin/java | sed "s_/jre/bin/java__") |
Save export for future reboots:
echo 'export JAVA_HOME=$(readlink -f /usr/bin/java | sed "s_/jre/bin/java__")' > /etc/profile.d/java_18.sh chmod +x /etc/profile.d/java_18.sh |
Download and install Maven:
wget http://apache.volia.net/maven/maven-3/3.3.9/binaries/apache-maven-3.3.9-bin.tar.gz tar -zxf apache-maven-3.3.9-bin.tar.gz; mv apache-maven-3.3.9 /opt/export PATH=/opt/apache-maven-3.3.9/bin:$PATH echo 'export PATH=/opt/apache-maven-3.3.9/bin:$PATH' > /etc/profile.d/maven.sh chmod +x /etc/profile.d/maven.sh |
On Ambari node install and enable docker (we will need it to build Metron mpack for Ambari):
yum install docker-io -y service docker start |
Also on your build box, install npm (needed to build metron-config, part of the UI):
yum install npm -y |
On main node clone Metron repository:
git clone https://github.com/apache/incubator-metron |
If you install Metron on single node (not multi-node as advised) you need to modify ElasticSearch config templates to use only configuration specified below. Config templates are located in incubator-metron/metron-deployment/packaging/ambari/metron-mpack/src/main/resources/common-services/ELASTICSEARCH/2.3.3/package/templates/
cluster.name: metron network.host: ["_eth0:ipv4_","_local:ipv4_"] discovery.zen.ping.unicast.hosts: [ <single_node_hostname> ] path.data: /opt/lmm/es_data index.number_of_replicas: 0 |
Fix Kibana install file (it should be fixed in METRON-641:
sed -i 's@{}/kibana@{0}/kibana@g' incubator-metron/metron-deployment/packaging/ambari/metron-mpack/src/main/resources/common-services/KIBANA/4.5.1/package/scripts/kibana_master.py |
Build Metron with HDP 2.5 profile:
cd incubator-metron mvn clean install -DskipTests -PHDP-2.5.0.0 cd metron-deployment/packaging/docker/rpm-docker mvn clean install -DskipTests -PHDP-2.5.0.0 |
On all nodes create localrepo directory and copy RPMs from Ambari node there:
mkdir /localrepo cp -rp /root/incubator-metron/metron-deployment/packaging/docker/rpm-docker/RPMS/noarch/* /localrepo/ |
Use scp for remote nodes:
scp /localrepo/* <replace_with_node_ip>:/localrepo/ |
If passwordless ssh has not yet been set up within the cluster, then in main node generate key:
cat /dev/zero | ssh-keygen -q -N "" 2>/dev/null |
Add this key to all the slave nodes:
ssh-copy-id -i ~/.ssh/id_rsa.pub <replace_with_node_ip> |
Inspired by: http://docs.hortonworks.com/HDPDocuments/Ambari-2.4.1.0/bk_ambari-installation/content/ch_Getting_Ready.html
Adjust limits to secure level (inspired by link):
ulimit -n 32768 ulimit -u 65536 echo -e "* - nofile 32768\n* - nproc 65536" >> /etc/security/limits.conf |
Enable time sync, disable firewall and SElinux:
chkconfig ntpd on service ntpd start chkconfig iptables off /etc/init.d/iptables stop setenforce 0 |
Make sure each node can resolve every other node's hostname or add hostname of each node to /etc/hosts on every node. For example add following lines in /etc/hosts of each node:
10.10.10.1 node1 10.10.10.2 node2 10.10.10.3 node3 |
Where 10.10.10.1, 10.10.10.2 and 10.10.10.3 are the IPs of your nodes and node1, node2 and node3 are hostnames.
On main node download and setup Ambari repo (you may replace the "2.4.1.0" with a newer Ambari version number):
wget -nv http://public-repo-1.hortonworks.com/ambari/centos6/2.x/updates/2.4.3.0/ambari.repo -O /etc/yum.repos.d/ambari.repo |
Check that it was added:
# yum repolist | grep ambari Updates-ambari-2.4.1.0 ambari-2.4.1.0 - Updates |
Install and setup Ambari server:
yum install ambari-server -y ambari-server setup -s |
Add Metron service to Ambari by running mpack command (make sure to specify correct path to mpack in --mpack=):
ambari-server install-mpack --mpack=incubator-metron/metron-deployment/packaging/ambari/metron-mpack/target/metron_mpack-1.0.0.0-SNAPSHOT.tar.gz --verbose |
Start Ambari:
ambari-server start |
Access the Ambari UI by going to the following URL in a web browser (use admin / admin as user / pass):
http://<replace_with_master_node_ip>:8080/#/installer/step0 |
Get Started page: Enter any desired cluster name.
Select Version: Make sure "Public Repository" is checked.
Install Options: Specify hostnames of your nodes where Ambari cluster should be installed (all the ones you have specified in /etc/hosts) in Target Hosts. Copy content of the main node private key (/root/.ssh/id_rsa) in "Host Registration Information". If you receive warning like below, ignore it and click OK:
The following hostnames are not valid FQDNs |
Choose Services: Select following Services:
HDFS YARN + MapReduce2 HBase Zookeeper Storm Flume Kafka Elasticsearch Kibana Metron Ambari Metrics |
Pig Tez Slider |
Assign Masters: Assign "Kafka Broker" on all nodes. Make sure move following components on one common node:
Storm UI Server Metron Indexing MySQL Server Kibana Server Elasticsearch Master Metron Parsers Metron Enrichment |
Assign Slaves and Clients: select All for:
DataNode NodeManager RegionServer Supervisor Client |
Customize Services: Following are the list of service that need to be configured:
Set zen_discovery_ping_unicast_hosts to: <replace_with_elasticsearch_master_hostname> (to the IP of the node where you assigned ElasticSearch Master on the Assign Master tab) |
Set kibana_es_url to: http://<replace_with_elasticsearch_master_hostname>:9200 (to the IP of the node where you assigned ElasticSearch Master on the Assign Master tab) |
Set Elasticsearch Hosts to: <replace_with_elasticsearch_master_hostname> (to the IP of the node where you assigned ElasticSearch Master on the Assign Master tab) Change global.json template from (unless it is already fixed as reported in METRON-642): "es.ip": "{{ es_url }}", to: "es.ip": "<replace_with_elasticsearch_master_hostname>", "es.port": "9300", |
Set rest of the configuration values to recommended by Ambari or the one you desire (like DB passwords) and perform install.
Fix ElasticSearch permission (it will crash right after start in Ambari) (unless it is already fixed as reported in METRON-642):
chown -Rh elasticsearch:elasticsearch /etc/elasticsearch |
As it will be owned by root by default and will drop error:
Likely root cause: java.nio.file.AccessDeniedException: /etc/elasticsearch/scripts # ls -la /etc/elasticsearch |
Fix path to ES log file in Java parameter (unless it is already fixed as reported in METRON-642)::
sed -i 's@elasticsearchelasticsearch@elasticsearch/elasticsearch@g' /etc/sysconfig/elasticsearch |
It is ok if some service will not able to start, check the errors and start them all manually.
Ignore the error Storm UI shown on the screenshot below if you've built your Metron code with HDP-2.5.0.0 profile (in Maven):
It appears because your Kafka Topic was not created or contains no data. Setup streaming and make sure your Kafka topic (from which topology should read data) exists.
If you have GUI installed on your server you should run following, before running git clone command:
unset SSH_ASKPASS |
Or you may receive an error:
(gnome-ssh-askpass:15028): Gtk-WARNING **: cannot open display: |
If Ambari metrics is not coming up, use this:
cd /usr/lib/python2.6/site-packages/resource_monitoring/ python psutil/build.py |
And re-try.
If you receive an error:
Unsupported major.minor version 52.0 |
You may need to install Java 1.8 on the node you receive this from:
yum install java-1.8.0-openjdk java-1.8.0-openjdk-devel -y |