2.8. Clustering and High Availability

Read this section for a summary of changes to clustering and high availability support and relevant configuration tools between Red Hat Enterprise Linux 6 and Red Hat Enterprise Linux 7.

2.8.1. Luci replacement limitations (pcs)

In Red Hat Enterprise Linux 6, luci controlled both Red Hat Enterprise Linux 5 and Red Hat Enterprise Linux 6 high availability clusters.
Red Hat Enterprise Linux 7 removes luci and replaces it with pcs. pcs can control only Red Hat Enterprise Linux 7 pacemaker-based clusters. It cannot control Red Hat Enterprise Linux 6 rgmanager-based high availability clusters.

2.8.2. Keepalived replaces Piranha

The Load Balancer Add-On for Red Hat Enterprise Linux 7 now includes the keepalived service, which provides both the functionality available in piranha and additional functionality. piranha is therefore superseded by the keepalived service in Red Hat Enterprise Linux 7.
As a result, the configuration file and its format have changed. keepalived is configured in the /etc/keepalived/keepalived.conf file by default. Details on the configuration format and syntax expected by this file are covered in the keepalive.conf man page:
$ man keepalived.conf

2.8.3. Online migration limitations

Online migration from Red Hat Enterprise Linux 6 to Red Hat Enterprise Linux 7 is not supported for clusters.
Additionally, the Red Hat Enterprise Linux 6 high availability stack is not compatible with the Red Hat Enterprise Linux 7 high availability stack, so online migration is not supported from a Red Hat Enterprise Linux 6 to a Red Hat Enterprise Linux 7 high availability cluster.

2.8.4. New resource manager (Pacemaker)

As of Red Hat Enterprise Linux 7, rgmanager and cman are replaced by pacemaker and corosync.
Pacemaker is a high availability resource manager with many useful features.
  • Detection and recovery from machine and application-level failures.
  • Support for many redundancy configurations.
  • Support for quorate and resource-driven clusters.
  • Configurable strategies for dealing with quorum loss (when multiple machines fail).
  • Support for specifying application startup and shutdown ordering, regardless of which machine the applications are on.
  • Support for specifying that applications must or must not run on the same machine.
  • Support for specifying that an application should be active on multiple machines.
  • Support for multiple modes for applications, such as master and slave.
  • Provably correct responses to any failure or cluster state.
  • Responses to any situation can be tested offline, before the situation exists.
For further information about Pacemaker, see the Red Hat Enterprise Linux 7 High Availability Add-On documentation available from http://access.redhat.com/site/documentation/Red_Hat_Enterprise_Linux/.

2.8.5. New feature: resource agents

Red Hat Enterprise Linux 7 introduces resource agents that work with the Pacemaker resource manager. Resource agents abstract cluster resources and provide a standard interface for managing resources in a cluster environment. For further information about the resource agents available in Red Hat Enterprise Linux 7, see the Red Hat Enterprise Linux 7 High Availability Add-On documentation available from http://access.redhat.com/site/documentation/Red_Hat_Enterprise_Linux/.
Support for IBM DB2 resource agents to drive and manage DB2 as cluster resource in High Available environments has been added in Red Hat Enterprise Linux 7.2.

2.8.6. Changed quorum implementation

qdiskd, as it was shipped in Red Hat Enterprise Linux 6, has been removed from Red Hat Enterprise Linux 7. The new quorum implementation is provided by votequorum, which is included in the corosync package, and which has been extended to replace qdiskd for most use cases. The extensions (wait_for_all, auto_tie_breaker and last_man_standing) are fully documented in the votequorum.5 man page.
$ man 5 votequorum