Red Hat HA Configuration

Latest response

I am working on a project that has a very limited budget and they want to use RHEL HA. They do not have centralized storage, all storage is local to servers. Is it possible to setup HA using nfs shares from each of the two servers that will be in the HA configuration? I know using external/centralized storage is the best but not sure if it is possible to purchase at this time.


Hi John,

What is the kind of data that you want both servers to access, is a database, web server, or anything else? What is the size of the data ?

The best way is to use a shared storage. If the data is not frequently changing, then you can let the data be on the local storage of both the servers and rsync them manually or through a cron job. You can also have a separate server to just act as a NFS server, which will also serve the purpose of a shared storage.

Swapnil Jain


Great question!

Just to clarify (and further Swapnil's comments above), are you wanting to use this cluster to serve files via NFS? or are you asking if NFS can be used for the quorum disk in lieu of shared (central) storage?

Which services do you plan to provide from your HA cluster of nodes?
How many nodes do you plan to configure in your cluster?

If your question regards the quorum disk, from my memory of using HA you can configure a cluster without a quorum disk using network only and access to an upstream router is used as the tie breaker. I will try and find a specific reference to this configuration and post the link.

This link may help answer your question

Specifically (emphasis mine):
Storage — Some type of storage is required for a cluster. Figure 1.1, “Red Hat High Availability Add-On Hardware Overview” shows shared storage, but shared storage may not be required for your specific use.

The cluster administration document index is here:

Someone from Red Hat may be able to give specific advice about supported configurations.

There will be two servers in the cluster. I will be running a Postgresql server and an ActiveMQ server on the nodes. The goal of this is to not have an interruption of service should a node fail and to not require manual intervention in the event a node fails. I would think there would be some sort of shared storage involved, especially when running a database.

If its database then you surely need a shared storage. But if you dont want to spend on that then theres still one way. Its "drbd". I have used it long back and it works very well. It creates a network block device, its like mirroring 2 partitions on different servers. So you can have /var/lib/pgsql mounted on the drbd partition before starting the service.

This should work. Good Luck!!

Swapnil Jain