How To Configure Production-Ready Mesos Cluster on Photon Hosts

In this post I will try to explain how to install a production ready Mesos cluser and Zookeeper. If you are not familiar with some of these technologies, you can read more about Mesos, Marathon and Zookeeper.


For this setup I will use 3 Mesos masters and 3 slaves. On each Mesos master I will run a Zookeeper, means we will have 3 Zookeepers as well. Mesos cluster will be configured with a quorum of 2. For networking Mesos use Mesos-DNS. I tried to run Mesos-DNS as container, but got into some resolving issues, so in my next post I will explain how to configure Mesos-DNS and run it through Marathon. Photon hosts will be used for masters and slaves.






Masters Installation and Configuration
First of all we will install Zookeeper. Since currently there is a bug in Photon related to Zookeeper installation I will use the tarball. Do the following for each master:

root@pt-mesos-master1 [ ~ ]# mkdir -p /opt/mesosphere && cd /opt/mesosphere && wget
root@pt-mesos-master1 [ /opt/mesosphere ]# tar -xf zookeeper-3.4.7.tar.gz && mv zookeeper-3.4.7 zookeeper

Example of the Zookeeper configuration file:

root@pt-mesos-master1 [ ~ ]# cat /opt/mesosphere/zookeeper/conf/zoo.cfg | grep -v '#'

Example of zookeeper systemd configuration file:

root@pt-mesos-master1 [ ~ ]# cat /etc/systemd/system/zookeeper.service 
Description=Apache ZooKeeper

ExecStart=/bin/bash -c "/opt/mesosphere/zookeeper/bin/ start-foreground"


Need to attribute the server id to each machine by creating a file named myid, one for each server, which resides in that server’s data directory, as specified by the configuration file parameter dataDir. The myid file consists of a single line containing only the text of that machine’s id. So myid of server 1 would contain the text “1” and nothing else. The id must be unique within the ensemble and should have a value between 1 and 255.

echo 1 > /var/lib/zookeeper/myid

Now lets install Mesos masters. Do the following for each master:

root@pt-mesos-master1 [ ~ ]# yum -y install mesos
Setting up Install Process
Package mesos-0.23.0-2.ph1tp2.x86_64 already installed and latest version
Nothing to do
root@pt-mesos-master1 [ ~ ]#

Example of master service systemd configuration file that should be located on each master:

root@pt-mesos-master1 [ ~ ]# cat /etc/systemd/system/mesos-master.service 
Description=Mesos Slave

ExecStart=/bin/bash -c "/usr/sbin/mesos-master \
	--ip= \
	--work_dir=/var/lib/mesos \
	--log_dir=/var/lob/mesos \
	--cluster=EXAMPLE \
	--zk=zk://,, \


Make sure you replace –ip setting on each master. Add server id to the configuration file, so zookeeper will understand the id of your master server. Should be done for each master with its own id.

root@pt-mesos-master1 [ ~ ]# echo 1 > /var/lib/zookeeper/myid 
root@pt-mesos-master1 [ ~ ]# cat /var/lib/zookeeper/myid 

So far we have 3 masters with Zookeeper and Mesos packages installed. Lets start zookeeper and mesos-master services on each master:

root@pt-mesos-master1 [ ~ ]# systemctl start zookeeper
root@pt-mesos-master1 [ ~ ]# systemctl start mesos-master
root@pt-mesos-master1 [ ~ ]# ps -ef | grep mesos
root     11543     1  7 12:09 ?        00:00:01 /opt/OpenJDK- -Dzookeeper.log.dir=. -Dzookeeper.root.logger=INFO,CONSOLE -cp /opt/mesosphere/zookeeper/bin/../build/classes:/opt/mesosphere/zookeeper/bin/../build/lib/*.jar:/opt/mesosphere/zookeeper/bin/../lib/slf4j-log4j12-1.6.1.jar:/opt/mesosphere/zookeeper/bin/../lib/slf4j-api-1.6.1.jar:/opt/mesosphere/zookeeper/bin/../lib/netty-3.7.0.Final.jar:/opt/mesosphere/zookeeper/bin/../lib/log4j-1.2.16.jar:/opt/mesosphere/zookeeper/bin/../lib/jline-0.9.94.jar:/opt/mesosphere/zookeeper/bin/../zookeeper-3.4.7.jar:/opt/mesosphere/zookeeper/bin/../src/java/lib/*.jar:/opt/mesosphere/zookeeper/bin/../conf: org.apache.zookeeper.server.quorum.QuorumPeerMain /opt/mesosphere/zookeeper/bin/../conf/zoo.cfg
root     11581     1  0 12:09 ?        00:00:00 /usr/sbin/mesos-master --ip= --work_dir=/var/lib/mesos --log_dir=/var/lob/mesos --cluster=EXAMPLE --zk=zk://,, --quorum=2
root     11601  9117  0 12:09 pts/0    00:00:00 grep --color=auto mesos

Slaves Installation and Configuration
Steps for configuring mesos slave are very simple and not different from masters installation. The difference we wont install zookeeper on each slave and will start mesos slaves in a slave mode and will tell the daemon to join the mesos masters. Do the following for each slave:

root@pt-mesos-node1 [ ~ ]# cat /etc/systemd/system/mesos-slave.service 
Description=Photon instance running as a Mesos slave,docker.service
ExecStartPre=/usr/bin/rm -f /tmp/mesos/meta/slaves/latest
ExecStart=/bin/bash -c "/usr/sbin/mesos-slave \
	--master=zk://,, \
        --hostname=$(/usr/bin/hostname) \
        --log_dir=/var/log/mesos_slave \
        --containerizers=docker,mesos \
        --docker=$(which docker) \
        --executor_registration_timeout=5mins \
        --ip=$(ifconfig ens160 | grep inet | awk -F'addr:' '{print $2}' | awk '{print $1}')"

Please make sure to replace the NIC name under –ip setting. Start the mesos-slave service on each node.

root@pt-mesos-node1 [ ~ ]# systemctl start mesos-slave

Now you should have ready mesus cluster with 3 masters, 3 zookeepers and 3 slaves.
Screen Shot 2015-12-24 at 2.22.27 PM

In my next post I will explain how to configure Mesos-DNS and Marathon.