Lohika Odessa Big Data #TechTalks

1. Docker based architecture

2. About Me
7+ years as backend Software Engineer.
● Java/Scala
● SQL/NoSQL databases
● distributed caches, queuing
● Hadoop/Spark
● AWS,DevOps
● REST, microservices

3. An approach to developing a single application as
a suite of small services, each running in its
own process and communicating with
lightweight mechanisms
http://martinfowler.com/
Microservice architecture style

4. • Component decomposition
• Independant development lifecycle and
technology freedom
• Better extensibility
• Fault isolation
Microservice benefits

5. • Deployment of much greater number of
services and components
• Monitoring of many services for performance
or issues
• Looking into log files from many micro-services
• Versioning between services and environment
consistency
Microservice drawbacks

6. 1. Multiple Service instances per host
● conflicting resource requirements(CPU, memory, IO, etc.)
● conflicting dependency versions
1. Single Service instance per host
● potentially less efficient resource utilization
● hosts provisioning and maintenance
1. Service Instance per VM
● building a VM image is slow and time consuming
1. Service instance per container
● The infrastructure for deploying containers is not as rich as
the infrastructure for deploying virtual machines
Microservice deployment patterns

7. Containerization vs Virtualization

8. Containerization providers
Implementations:
● by vendors(Open Source,HP, IBM, Microsoft)
● by OS (Linux, FreeBSD, Solaris, Windows)
http://en.wikipedia.org/wiki/Operating-system-level_virtualization#Implementations.
Docker has become the most popular nowadays.

9. Docker Ecosystem

10. How it works
http://en.wikipedia.org/wiki/Docker_(software)

11. Docker registry

12. Docker container overhead

13. Docker container network overhead

14. Basic docker usage scenario
1. create Dockerfile
2. build image with specific name in format
[registry_url/]repository:imageTag , e.g
myhost:5000/mycompany/myservice:version1
1. push image into registry (private or Docker Hub)
2. pull image by repository and tag (Optional)
3. run container from image

15. Dockerfile
• inherit existing images
• add/copy files to container
• set environment vars
• run shell command
• expose ports
• set working dir
• share host FS with container
• --no-cache option
• CMD and bootstrapping
• CMD vs ENTRYPOINT
FROM mycompany/app-base:latest
RUN mkdir -p /opt/app
ADD core.jar /opt/app/core.jar
RUN curl http://3rd-paty.repo.org/download/component.zip &&
unzip component.zip -d /opt/app/ && rm -f component.zip
COPY lib/ /opt/app/lib
ADD bootstrap.sh /etc/mysevice/
RUN chmod +x /etc/mysevice/bootstrap.sh
ADD workspace-manager-core.jar
ADD logback.xml /opt/app/logback.xml
ENV MODE="PRODUCTION"
VOLUME /log
WORKDIR /opt/app/
EXPOSE 80
CMD /etc/myservice/bootstrap.sh && /usr/lib/jvm/latest/bin/java
-cp /opt/app/core.jar:/opt/app/lib
-Xms256m
-Xmx2048m
-XX:+UseCompressedOops
-XX:+UseG1GC
-Dspark.kryoserializer.buffer.max.mb=128 -
Dlogback.configurationFile=/opt/app/logback.xml
org.mycompany.app.Runner

16. Ready, set,go
• docker build -t mycompany/app:1.0
• docker tag -t mycompany/app:1.0
registry.mycompany.com:5000/mycompany/app:1.0
• docker push
registry.mycompany.com:5000/mycompany/app:1.0

17. Container entrypoint
• Container must have long leaving process -
entrypoint
• Container will become stopped when process is
ended/crashed
• Two strategies: process per container vs
supervisord

18. Supervisord configuration
[supervisord]
nodaemon = true
logfile = /var/log/supervisord.log
[program:hadoop-hdfs-namenode]
command = /etc/init.d/hadoop-hdfs-namenode start
[program:mesos-master]
command = /usr/local/sbin/mesos-master --work_dir=/mnt/mesos-work --
hostname={{ MESOS_HOSTNAME }} --log_dir=/var/log/mesos
[program:vw-spanning-tree]
command = /usr/local/bin/spanning_tree

19. FROM mycompany/mesos-general
MAINTAINER mycompany
#Configure namenode
RUN apt-get install -y hadoop-hdfs-namenode=2.5.0+cdh5.3.3*
#Add config templates
ADD hadoop-conf/core-site.xml /etc/hadoop/conf/core-site.xml.tpl
ADD hadoop-conf/hadoop-env.sh /etc/hadoop/conf/hadoop-env.sh
ADD hadoop-conf/hdfs-site.xml /etc/hadoop/conf/hdfs-site.xml.tpl
ADD bootstrap.sh /bootstrap.sh
RUN chmod +x /bootstrap.sh
COPY supervisord.conf /etc/supervisor/conf.d/supervisord.conf.tpl
CMD /bootstrap.sh && /usr/bin/supervisord
Master Dockerfile

20. #!/bin/bash
set -e
mkdir -p /mnt/mesos-work
mkdir -p /var/log/mesos
source /resolve_ec2_hostname.sh
envtpl < /etc/hadoop/conf/core-site.xml.tpl > /etc/hadoop/conf/core-site.xml
envtpl < /etc/hadoop/conf/hdfs-site.xml.tpl > /etc/hadoop/conf/hdfs-site.xml
envtpl < /etc/supervisor/conf.d/supervisord.conf.tpl > /etc/supervisor/conf.d/supervisord.conf
Bootstrap.sh example

21. • Container identification (--name)
• Network settings
o --net = [none, bridge (default), host, container:<name|id>]
o -P vs -p
o --link
• Useful options for debug/testing
o --rm=true, -it
o docker exec -it running_container_name bash
• Env variables
• Runtime constraints on CPU and memory
Run options

22. Environment example

23. Thank You!

24. References
● http://martinfowler.com/articles/microservices.html
● https://docs.docker.com/introduction/understanding-docker/
● http://domino.research.ibm.com/library/cyberdig.nsf/papers/092905219
5DD819C85257D2300681E7B/$File/rc25482.pdf