The Art of Clustering Dinos and Elephants – ClusterLabs Summit in Nuremberg 2017
The following article has been contributed by Tanja Roth, Technical Writer at the SUSE Documentation Team.
Communication between cluster nodes is key in High Availability clusters. The same is true for persons working on this topic. Therefore, community members and HA experts from around the globe meet in regular intervals at the ClusterLabs summit to discuss the whereabouts of HA clustering in the open source ecosystem.
More than 10 years ago, the evolution of HA clustering in an open source environment started off with different approaches and solutions. Nowadays basically all open source HA clusters are based on Pacemaker as the Cluster Resource Manager and Corosync as the messaging and membership layer.
While the last summit had taken place 2015 in Brno, Czech Republic, this year’s summit on Sep 6-7 was located at the SUSE headquarters in Nuremberg, Germany. About 60 participants from different vendors and Linux distributors like Alteeve, Citrix, LinBit, SUSE, and Red Hat found their way to Nuremberg, thus making it the biggest HA community summit so far.
The event was organized by Kristoffer Groenlund, based in Sweden, and Chris Kowalczyk. The agenda covered a wide range of topics, for example:
- changes in the underlying bits and pieces of the HA stack (that administrators and users usually don’t get to see)
- tools and interfaces used by cluster administrators and operators
- storage solutions
- the use of Pacemaker clusters in OpenStack
On Wednesday morning at 9:15 AM Pacemaker creator and main contributor Andrew Beekhof started us off with a talk about the container ‘bundle’ feature in Pacemaker.
Other talks on that day included topics like future plans for SBD (Storage Based Death), ‘Intelligent Availability’, an introduction to openATTIC, and multi-site (or geographically dispersed) clusters.
In the evening, we had dinner at a restaurant in the old town of Nuremberg. Soon the large room was filled with lively banter and laughter. Some participants continued the technical discussions started during the day, others just enjoyed the informal conclusion of the day with colleagues and community members from different parts of the world.
Day two of the summit was kicked off by Chrissie Caulfield who updated us on Corosync 3 and Kronosnet (knet), followed by a talk about the pcs command line tool. In the afternoon, Adam Spiers presented the usage of HA in OpenStack. Fabian Herschel and Frank Danapfel demonstrated how they integrate Pacemaker clusters into SAP solutions like SAP NetWeaver and SAP HANA.
Beside the presentations, the schedule also gave enough room for discussions. Although some tools are distribution-specific (‘pcs’ vs. ‘crm shell’ command line tool, ‘Hawk2’ vs. ‘Picus’ web interface), a lot of productive discussions emerged between the community members. At the end of day two, 60-90 minutes were dedicated to collecting and discussing ideas about the future development of Pacemaker clusters. We also reached quorum about the date for the next ClusterLabs summit, which is to take place in 2019. However, the location was not determined other than to agree that it should probably be somewhere in Europe again.
From my point of view, it was a prolific summit and a great opportunity to talk to community members from Australia, Canada, China, Sweden, France, England, Germany and the Czech Republic (to name but a few).
While there is some variety in the tools used and developed by different vendors (and some differences in the company cultures), there was a lot of agreement, too. After all, collaboration and community-wide support of a common cluster stack helps all our customers to protect their mission-critical workloads and services with a reliable and stable open source HA solution.