High Availability

From SME Server
Revision as of 19:39, 23 January 2013 by RequestedDeletion (talk | contribs) (→‎Technical overview)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigationJump to search
PythonIcon.png Skill level: Developer
Risk of inconsistencies with Koozali SME Server methodology, upgrades & functionality is high. One must be knowledgeable about how changes impact their Koozali SME Server. Significant risk of irreversible harm.


!!! WORK IN PROGRESS, DO NOT USE !!!


Important.png Note:
Do not apply this how-to to any production servers!! Plan your cluster and HA capabilities carefully and test extensively on virtual servers


High Availability on SME8

Introduction

The clustering technique provides many possible scenario's depending on the requirements. This could be an active/active 16-node cluster or a simple 2-node active/standby failover cluster. This How-To is will only address and focus on a High Available 2-node active/standby cluster, referred to as "SME-cluster".

CentOS6.3 (and RHEL6) still use the same components, so the know-how you will gain from this How-To will be applicable to a possible new release of SME server based on CentOS6.3. Next to that, the most valuable know-how is about understanding and configuring the cluster and HA capabilities, which will be conceptually the same for any cluster environment. Especially since some components (dependencies) like openais, soon to be replaced by corosync (as of RHEL6), are the de facto standards for Linux/Unix. RHEL has no know plans to change this strategy for any future RHEL release.

Clustering and HA is relatively easy to implement for there is a lot of knowledge and how-to's available from the RHEL and CentOS communities. Next to this, clustering is not a new technologie and has been around for quite some time as a proven technology. The hard part is precise planning and configuring correctly.

Disclaimer

Use at your own risk. The authors/contributers nor contribs.org are responsible or liable for any damages as a result of using this information. It is provided as is.

Technical overview

The main components we will use in this how-to are based on the default clustering components of CentOS 5.8. These include

  • ricci, the messages manager
  • cman, the cluster manager
  • rgmanager, the services manager
  • drbd, the device block driver

and their dependencies which will be installed automatically by yum. The compete list of packages that will be installed is:

TBA

There is also a webgui component called luci, but this will not be used for it is not working correctly on SME8. Configuration of the cluster will be done with the tools that come with the above applications.

Planning

Business Adaptation

Installing

Configuration

Testing

Production