Red Hat CLUSTER SUITE FOR ENTERPRISE LINUX 5.1 Manual do Utilizador descarregar pdf (Página 25)

Introduction to Linux Clustering

8 Cluster Examples and Challenges

There are numerous ways you can configure a cluster, which will depend on your requirements and

budgets. There are also some complex requirements if you wish to have virtualization with the

cluster as well as building geographically separated clusters.

This document details a number of examples of cluster designs that may be suitable for you and to

just give you an understanding of what is possible as well as discussions of the problems and

limitations with each design.

8.1 Two node HA cluster with DRBD

A common high-availability requirement is to make a particular server survive any hardware

failure. The solution to this, is to add a second identical server and setup a HA cluster

between them.

To make a two-node high-availability cluster work, we have the following set of

requirements:

● Data must be accessible by both servers, with both servers being able to read/write at

the same time.

● If an individual service die, it should resume on the secondary server.

● If one server suffers a complete failure, the secondary server should resume all tasks.

Solution:

● Both servers identical hardware, running Redhat Cluster Suite.

● Local root and swap filesystems, remaining disk space turned into DRBD block

device setup with both nodes running as primary.

● DRBD device running GFS with journal space for 2 nodes and both nodes mounted

at the same time.

● Each service configured in cluster suite with a floating IP address. Any service that

fails will relocate to the second node.

● In the event of a full server failure, the second node will resume all services.

Notes:

● DRBD is used primary/primary with GFS so that both servers can be running

services at once. This may be undesirable or unwanted if you only have one floating

IP address, in this scenario you need to group all the services using that IP together.

In that case, it would also be okay to run DRBD as primary/secondary with a

traditional filesystem like ext3 or xfs and have all the services configured as a single

resource, to all failover together.

1 2 ... 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Sem comentários

Red Hat CLUSTER SUITE FOR ENTERPRISE LINUX 5.1 Manual do Utilizador Página 25