OpenStack foundation holds ‘OpenStack Summit Austin 2016’ from April 25 to 29 at Austin Convention Center in Austin, Texas. In this event, recent OpenStack version ‘Mitaka’ and various cases of application related to OpenStack are introduced.

‘OpenStack Summit Austin 2016’ from April 25 to 29 at Austin Convention Center in Austin is an event where more than 7,500 participants from more than 55 countries are attending. In this global event, enterprise IT strategy, cloud application development, telecom ecosystem and NFV architecture aspects, and new projects and technologies in community are covered. Also, at developer’s working session, it will be discussing roadmap of ‘Newton,’ which is the next release on probably October 2016.

In this summit, Ancestry (Ancestry.com) introduced about its OpenStack and the application case of Kubernetes through ‘Ancestry.com in Production with OpenStack and Kubernetes’ session. Its cases had been introduced in the last Tokyo summit, but the cases in this summit include few latest news. In this session, they introduced about detailed application method of its Swift, Kubernetes, CoreOS, etc.



Ancestry used a solution of SwiftStack in application of Swift, because of the reasons such as having professionalism, as it is positioned at the largest contributor in the community, deployment experience of multi-petabyte Swift environment, controller, and technology support aspect. Swift’s Amazon S3 API emulation is mentioned as an unusual point of Ancestry’s environment. This is caused by hybrid composition with Amazon service, and it is introduced that it enables to use both sides through common API.

Ancestry introduced that it used Avere gateway hardware as a solution of connecting CIFS, NFS, etc. that is required from a specific software and Swift environment. An interesting point is that flash-mirror and flash-move functions are used at data synchronization between on premise environment and public cloud. In the future, the tasks for this would be add key synchronization and cloning functions between swift and S3, closer integration between private and public cloud, etc.

In aspect of OpenStack, it is introduced that ‘new culture’ is needed as the existing specific vendor base to open source base, and its examples would be self-service portal based on horizon, DevOps model, which integrates management and development closely, a part that requires understanding about a broad part, automation, etc. Also, the advantages by using OpenStack are shorter time to market by rapid test progression, saving provisioning time through cloud optimized image, and the existence of self-service portal.

There are five things to successful deployment of OpenStack environment. The first thing is the aspect of understanding about OpenStack cluster’s deployment and test, and whether doing deployment by oneself or by getting professional vendor’s support should be considered. In case of composing by oneself, the one can select only the essential thing and it is one of the important factors. Apart from this, automation, and solutions to the problem about new parts that we had never met are introduced.


Based on CentOS 7.1 and RDO Kilo version, Ancestry’s OpenStack environment is composed of the factors, such as Nova, Glance, Hit, Horizon, and Keynote, Midokura for network overlay, SwiftStack for object storage, etc. With high availability of 77-node cluster using Del R630 system, Nova used a composition that uses a local disk of compute node. Moreover, this whole cluster is divided into two racks depending on purposes.

The first rack for development and ELK is composed of one 3 node RMQ cluster, one 3 node ZooKeeper, Cassandra cluster, two HA Midokura BGP gateway nodes, 33 Nova compute nodes and three HA controllers for Kubernetes Canary. Moreover, 100 ElasticSearch containerized clusters and 25 VM Kibana Logstash containerized clusters are operated in CoreOS 991.2 beta. The second rack has a composition of 33 Nova compute nodes, DevStack and OSA for development and secondary user VM.

Distribution uses Python script, Ansible and PXE booting at the same time, and Bash script is used for automatically adding compute node to Mido tunnel zone. Also, after going through several settings, it took less than 20 seconds to distribute 75 ELK VMs and make them in the operating states. At the same time, the use of rapid drive for DB and recent version of Rabbit, file descriptor with higher than 65K, and cautions to the positions of setting and storage mount, and CoreOS bug are mentioned as advices.


The advantages of CoreOS, which Ancestry uses as a standard environment, are introduced; it is light, it has composition for container environment with strong security, it is frequently updated and convenient for management. In addition, internally, Kubernetes environment is deployed on three positions. Among them, it is deployed as DR cluster at AWS and as the purpose of development environment at OpenStack base. At production environment, bear metal environment, which has characteristics of overhead minimization and single overlay, is used. The existing complicated network setting is simplified, and a feature of this setting is that it uses NodePort at service approach which is more convenient way, and the pot’s IP is not routable. At monitoring of Kubernetes, Heapster collects and shows cAdvisor information, and visualizes information collected by Kubedash. At the same time, it is added that open source monitoring system ‘Prometheus’ has an advantage of integration with Grafana or time based database.

According to the service requirement, an interesting thing in the flaw of Kubernetes is that Jenkins and CoreOS are attached. It enabled to simply resolve when retrieving, updating, saving and changing the images into requirements. Through these points, it was possible to achieve successes such as reduction in distribution time or easier application updates by rolling updates or distribution.

Copyright © acrofan All Right Reserved