Linux-HA

1
Linux-HA

• High-Availability Linux

2
Concepts of Linux-HA

• Reliability
• Availability
• Serviceability

3
HA Clusters

• What are HA clusters?
• High-availability clusters are groups of
  computers that support sever applications that
  can be reliably utilized with a minimum of
  down-time.
• Utilizes redundant computers in clusters that
  provide service when any system components fail.

4
HA Clusters (cont.)

• Failover A process by which HA clusters detect
  hardware or software faults and restart the
  application on another system without requiring
  administrative intervention
• Emphasis on a layered approach to redundancies

5
Heartbeat Features

• Primary software of Linux-HA is called Heartbeat
• No fixed limit on nodes, allowing use with
  clusters of any size
• Parallel resource monitoring as with normal
  computing, but can shift resources from one node
  to another if the initial node fails
• Automatically removes failed nodes from the
  cluster

6
Heartbeat Features (cont.)

• Integrates with many popular software packages,
  including Apache, DB2, Oracle, PostgreSQL
• GUI included for easier controlling and
  monitoring of the clusters and relevant resources

7
Initial Designs

• Originally capable of only handling two nodes at
  a time
• Did not include resource monitoring
• Would later switch to a layered design
  implementing n-node clusters

8
Version 2.1.4

• Project was split into various separate packages
• Pacemaker Cluster resource manager component
  that handles resource management and node failure
• Heartbeat Now only refers to the layer used for
  communication between clusters and the individual
  nodes of the cluster.

9
Version 2.1.4

• Resource Agents a standardized interface for a
  cluster resource, used for translating operations
  to a cluster and determining success or failure
  per process
• Cluster Glue a set of libraries, tools, and
  utilities for use with Heartbeat and Pacemaker
  this includes everything not covered by Heartbeat
  (messaging), Pacemaker (resource management),
  Resource Agents (cluster operations)
• Local Resource Manager similar to the
  Pacemaker, but solely exists for one client in
  the cluster, and is thus not aware of the
  status of the rest of the cluster