Using DRBD to Build High-Availability Clusters on Commodity Hardware

Charlie Schluting of LINBIT will explain how DRBD works and how people currently use it, with enough information to get you started building your own clusters.

DRBD stands for Distributed Replicated Block Device, and as the name implies, allows you to replicate block devices over TCP. DRBD is extremely flexible due to the fact it is a block device, and as such is used in a variety of situations. At the most basic level, you can replicate data between two servers to provide synchronously replicated storage redundancy for either failover or disaster recovery purposes. In active/active mode, you can also run GFS, OCFS2, or other clustered file systems.

Topics that will be covered: - How it works, history, and future exciting news regarding mainline kernel inclusion - How it is used: HA-iSCSI, HA-NFS, Virtualization, Apache, Samba, etc. - Cluster Resource Manager options and recommendations, and news about the confusing changes in the Linux-HA / Clusterlabs communities.

And the majority of the time will be spent on: - Example cluster configuration: hardware setup, installation and configuration, and cluster manager integration.

Come with questions!

Brian Martin will be test-driving his IEEE presentation on his experiences in a true, "abandon the building" disaster recovery effort. He'll place particular emphasis on where technically sound, well-tested disaster plans often fail in a real disaster, and how these problems can be overcome.

The best data center disaster recovery plans are developed carefully and tested regularly. If you're at that stage, you may think you are well prepared. In this entertaining presentation, Brian Martin describes the unexpected problems that arose when a well thought out and tested plan met a real disaster, and draws lessons that are applicable to any disaster recovery situation.

Brian Martin has spent 30 years in the IT field, fairly evenly divided between being a mainframe system programmer and a server system administrator. He has operated Martin Consulting Services in the Portland Oregon area since moving to Portland from the San Francisco Bay Area in 1996. He has a wife, two cats, a dog, and nine in-service computers at home.

DRBD stands for Distributed Replicated Block Device. Mainline in the Linux kernel since 2.6.33, it is used to replicate data at the block level over the network in a "network RAID1" fashion. It is generally deployed as a cost effective, shared-nothing alternative to a SAN and used as the building block for high availability clusters. Pacemaker is currently the de facto open-source cluster resource manager (CRM) for Linux HA clustering. With it, nodes and services can be monitored and managed to ensure maximum uptime in the face of the most severe service and hardware level failures. Combining the two allows admins to %99.999 uptime at a fraction of the price of proprietary alternatives.

In LINBIT's second PLUG Advance Topics installment, Adam Gandelman will give a more in-depth view of DRBD and Pacemaker and demonstrate how they work closely together to keep applications running and consistent. During the second half of the presentation, Adam will provide attendees with a real-world example by configuring a highly-available LAMP cluster from the ground up. Though geared toward web services, the concepts presented can easily be expanded to provide the HA gaurantee to virtually any Linux service.

Agenda: - Brief re-introduction to DRBD, Pacemaker and HA clustering concepts. - Overview of various use cases and interesting deployments - Configuration and implementation of a highly-available LAMP cluster using DRBD for data redundancy and Pacemaker for resource management.

Adam Gandelman is an expert in open-source clustering and high availability. Originally from New England, Adam lives in Portland, OR where he has been working at LINBIT, developers of DRBD and maintainers of Heartbeat. Aside from providing top-level Linux High-Availability and Disaster Recovery consulting for customers in the Americas, he leads LINBIT training courses in the US, doubles as a technical writer and regularly contributes to related open-source projects.

If you've heard of Erlang, perhaps you've also seen the acronym "OTP". What the heck is "OTP" and why do the Jedi refer to it as the "magic" that gives Erlang it's reputation for scalability and fault tolerance?

Why is it called Erlang/OTP? What gives? How does this relate to Elixir? What should I know about it?

Nathan Aschbacher of Visa and Elixir Games PDX has kindly agreed to join us and share his wisdom, give us an overview and also, to levitate R2D2.

Our generous hosts this month are HouseHappy.

October we followup on this deep topic with Jeff Weiss showing self-healing application magic in Elixir and OTP. Hope to see you there!

• Steve

If you've heard of Erlang, perhaps you've also seen the acronym "OTP". What the heck is "OTP" and why do the Jedi refer to it as the "magic" that gives Erlang it's reputation for scalability and fault tolerance?

Why is it called Erlang/OTP? What gives? How does this relate to Elixir? What should I know about it?

Nathan Aschbacher of Visa and Elixir Games PDX has kindly agreed to join us and share his wisdom, give us an overview and also, to levitate R2D2.

Our generous hosts this month are HouseHappy.

October we followup on this deep topic with Jeff Weiss showing self-healing application magic in Elixir and OTP. Hope to see you there!

• Steve

Jeff Wiess will share his presentation from ElixirConf EU and give us a live demo:

Chat Bot: A Practical Walkthrough of the powerful Features Elixir/Erlang/OTP

Bring your laptop and be ready to help us try and crash Jeff's chat server service.

Written in Elixir, a "Ruby-like" flavor of Erlang this demo will show supervision trees, clustering and live code updating.

Our generous hosts this month are HouseHappy.

Erlang On NixOS - Managing And Releasing Erlang Systems In The Cloud With A Fully Declarative Package Manager

In this talk we will discuss how to manage Erlang dependencies with the Nix package manager and how to use the Nix system to deliver declaratively described images containing an Erlang Release to cloud platforms.

Talk objectives:

To educate the audiance about the value of using a functional, declarative package management system to deliver functional, declarative systems.

Target audience:

Developers actively deploying Erlang systems and those interested in deploying Erlang System.

About Eric Merritt

Co-author of Erlang and OTP in Action, open source contributor, Erlang Engineer.

Jesse Cook will lead our exploration tonight.

Jesse: A beginner Alchemist who's really enjoying the functional nature of Elixir and the design of the language.

Description: Moving complexity around - What's the best way to provide a unified API in front of some of the worst APIs out there? The tools I reached for are Phoenix, Absinthe for GraphQL and a series of adapters. These adapters utilize both polymorphism and metaprogramming, but was this the right way to do it in Elixir? Let's discuss the overall architecture and the nitty gritty details.