High Availability (HA) on Linux just celebrated its 10-year anniversary. For over a decade, free and open source software has been powering highly available services on commodity hardware. Over that time, the Linux high availability community has grown and matured, and its various projects are thriving. HA systems on Linux now rival or exceed the capabilities of proprietary systems.
The application-agnostic Linux high availability stack adds HA to services as diverse as MySQL and PostgreSQL relational databases, file services such as Samba and NFS, and firewalls and routers, to name but a few. The stack is a key component in both private and public cloud infrastructures, as well as the conventional data center.
This miniconf aims to bring together developers and users of HA and distributed storage software on Linux, including but not limited to:
- Pacemaker (cluster resource management)
- Corosync/OpenAIS, Heartbeat (cluster membership/messaging)
- OCFS2, GFS2, DLM (clustered filesystems, lock management)
- CTDB (clustered Samba)
- DRBD, GlusterFS, Ceph (distributed storage)
- SQL and NoSQL databases
- Caching layers (Redis, memcached, etc.)
If your system absolutely must be up all the time, or your data needs to be reliably available everywhere, this is the miniconf for you.
- 10:30 Storage Replication in High-Performance High-Availability (HPHA) Environments -- Florian Haas
- 10:55 Building a Non-Shared Storage HA Cluster with Pacemaker and PostgreSQL 9.1 -- Keisuke Mori
- 11:20 Break
- 11:30 Extend Pacemaker to Support Geographically Distributed Clustering -- Tim Serong on behalf of Jiaju Zhang
- 11:55 HiPBX - HiAv VoIP with Open Source Software and 5000 Lines of Bash -- Rob Thomas
- Live demonstration of HiAv phone system, bring a SIP Client or an Android phone to have a play with it too.
- 12:20 Lunch
- 13:20 Squashing SPOFs with Common Sense, Velcro, and a Hammer -- Rob Thomas
- 13:45 CTDB Overview -- Ronnie Sahlberg
- 14:10 Break
- 14:20 High Availability Login Services with Samba4 Active Directory -- Kai Blin
- 14:30 HA Lessons Learned from Darth Vader -- Ronnie Sahlberg
- 14:45 MySQL for the Developer in a Post-Oracle World -- Adam Donnison
- 15:10 Afternoon Tea
- 15:40 MySQL and Postgres Cloud Offerings -- Stewart Smith & Selena Deckelmann
- 16:05 Scaling Data: Postgres, The Stack and the Future of Replication -- Selena Deckelmann
- 16:30 Break
- 16:40 Swift 101 -- Monty Taylor
- 17:05 MySQL Web Infra Scaling and Keeping it Online, Cheaply -- Arjen Lentz
Storage Replication in High-Performance High-Availability (HPHA) Environments
This presentation highlights a new component in the HA stack: Flashcache.
Flashcache, originally developed at Facebook, is a generic block-level cache device that adds SSD caching capability to ordinary hard drives. While owing its origins to a desire to scale MySQL/InnoDB configurations, it is a generic caching layer that adds efficient caching to both database and non-database workloads.
In this presentation, we will examine new features in the high-availability Linux cluster stack and DRBD that make Flashcache suitable for hot failover in HA clusters, allowing us to add a healthy performance boost to cache-sensitive services that would otherwise take hours or days to recover from failover, until running at peak performance again.
For attendees, familiarity with DRBD and Pacemaker is a plus but not required. Prior knowledge of Flashcache is not necessary.
Florian is a Linux high availability specialist, experienced technical consultant, seasoned training instructor, and technical writer. He frequently contributes to projects like DRBD, Pacemaker, Heartbeat, Corosync, and others. Florian is the principal author of the DRBD User's Guide, the Linux-HA User's guide and a number of Technical Guides on high availability, and a sought-after conference speaker.
Building a Non-Shared Storage HA Cluster with Pacemaker and PostgreSQL 9.1
As of PostgreSQL 9.1 it provides a support of synchronous streaming replication, which will allow users to build a non-shared storage HA cluster. In order to execute an automatic fail-over when a failure occurred, integration with a HA cluster software such as Pacemaker is necessary.
We have developed an enhancement of 'pgsql' resource agent to enable the PostgreSQL streaming replication feature as a master-slave resource for Pacemaker. In this talk we will introduce a brief explanation about the resource agent, how to use it and how it works.
Keisuke MORI is a chief engineer at NTT DATA Intellilink Corporation and works for NTT OSS Center to provide engineering supports of Pacemaker, Heartbeat and the Linux-HA products for their customers. He is a contributor to maintain Pacmaker-1.0.* stable releases. He is also involved the Linux-HA Japan community (http://linux-ha.sourceforge.jp/) to promote Pacemaker and other HA products into Japanese users.
Extend Pacemaker to Support Geographically Distributed Clustering
This presentation will give an introduction on how to extend Pacemaker to support geographical distributed clustering. We have developed a so-called cluster ticket manager, named as Booth, to manages certain resources in the Geo-cluster environment.
Booth implments a Paxos algorithm which is used to guarantee distributed consensus. However one development goal of that is to make the paxos implementation in booth relatively independent and pluggable, which can also be extended to different paxos varients and thus be used in different using scenarios in the distributed storage field.
Currently the presenter works for SUSE Linux Products GmbH, maintains Corosync and DLM, co-maintains OCFS2 and some other packages for SUSE Linux Enterprise Server. He is the author of the Booth cluster ticket manager.
HiPBX - HiAv VoIP with Open Source Software and 5000 Lines of Bash
Rob Thomas will be presenting a live demonstration of Highly Available Asterisk, using HiPBX. This demo will show you how easy it is to build and use a HiAv phone system in your office, or for your clients. HiPBX is based on Scientific Linux 6 (But works on RHEL/CentOS 6), FreePBX 2.9, and uses the standard Linux-HA corosync, pacemaker, and drbd packages
Squashing SPOFs with Common Sense, Velcro, and a Hammer
Rob was the original Lead Dev of FreePBX (php/perl based GUI for Asterisk), sole (sigh) dev of HiPBX. Founder of SLUG, geek, motorbike enthusiast. I'm 'xrobau' everywhere (gmail, github, cpan, twitter, etc).
A short informal presentation about CTDB and how we use CTDB together with GPFS to create a clustered NAS server.
Conceptual overview of how the database works.
Mathematician by training, OSS hacker by choice. Have been hacking OSS since late 80s
Currently hacking on CTDB, WIRESHARK, STGT, DBENCH, iSCSI, KVM/QEMU, ...
High Availability Login Services with Samba4 Active Directory
If you are in an environment where you need to provide login services for a variety of client operating systems, Samba4 Active Directory might be a good fit. And because it's always painful when the login server goes down, Samba4 AD supports a distributed setup with multi-master replication that allows your clients to keep signing in even if one of your Domain Controllers goes down.
This presentation will give a quick overview of the replication features that allow a multi-DC AD setup with Samba4 without requiring an underlying cluster management software.
Kai Blin is a computational biologist by trade and Open Source developer by passion. Being more of a network and systems programmer in his spare time, the Samba Team member feels lucky to be able to also work on Open Source software on his day job. During the day he is helping to identify new antibiotics to combat diseases, both at the computer and in the lab.
MySQL for the Developer in a Post-Oracle World
MySQL is advertised as the most popular open source database in the world, even though at the time of Oracle's pending acquisition this was challenged by none other than Oracle's CEO, Larry Ellison. But beyond the corporate marketing, is MySQL still a viable alternative for database development for Open Source developers? In this talk I intend to walk through the recent developments in MySQL and how the players are lining up, and what this means for developers. Looking at competing databases like MariaDB and Drizzle and the companies that supply and support them.
Adam is currently the Webmaster at SkySQL.com, coming off of more than 5 years developing the web assets for MySQL AB through its most turbulent years and through acquisition by Sun Microsystems and then Oracle Corporation. In his position Adam has been involved in early testing of MySQL Server, MySQL Enterprise Monitor and MySQL Proxy and built high availability websites on top of MySQL technologies.
Scaling Data: Postgres, The Stack and the Future of Replication
Postgres has long been a database that easily scales vertically. It has also supported durable and reliable binary replication.
In the last two years, work on HA, clustering and extensions to our MVCC system has produced clustering features that help Postgres efficiently scale across many smaller systems.
This talk will start with a demo of the latest Postgres replication features. Then, we'll walk through open source and commercially-supported stacks that supporting the most demanding applications backed by Postgres.
In version 9.1, Postgres added streaming and synchronous replication that can be configured in about 5 steps. In 9.2 (to be released in 2012), we're adding cascaded replication and optimizations for common developer requests.
These features set the stage for a fast, efficient and distributed SQL cluster -- one that has configurable durability, is easy to administer and supported by "cloud" and virtualization vendors.
Selena Deckelmann is COO and founder of Prime Radiant. She's a major contributor to PostgreSQL. She founded the Portland PostgreSQL Users Group in 2006.
She's created three successful conference series about PostgreSQL, open source development and culture. She speaks internationally about free software, developer communities and trolling. Her interests include opening up government data with the City of Portland, urban chickens and finding ways to make databases run faster.
She founded and chairs Postgres Open, a conference dedicated to the business of PostgreSQL and disruption of the database industry. She founded and co-chaired Open Source Bridge, a developer conference for open source citizens. She founded the PostgreSQL Conference, a successful series of east coast/west coast conferences in the US for PostgreSQL. She's helped run other conferences and events like IgnitePDX, IgniteGov, WhereCampPDX, BarCampPDX and PG Days.
You can find her on twitter at @selenamarie.
MySQL Web Infra Scaling and Keeping it Online, Cheaply
Of course it's fun to talk about fancy gear, but it's not always feasible to use. In this session, Arjen shares how Open Query helps its clients with their infrastructure, given the real world realities of time and budget constraints. We'll discuss infrastructure design for both resilience (our word for "high availability") as well as scaling - often deployed in a step-wise manner over time.
Arjen is the founder and exec.director of Open Query, which provides remote maintenance for MySQL and related environments. Prior to that he was employee#25 at MySQL AB (2001-2007)