Setup Two Node Red Hat Cluster without shared storage
Hi,
We have plan to build a high availability critical application production, we need a suggestion and best practice to implement high availability server using red hat. Let say there are two node servers (two rack servers) connected each other thru network switch using their dedicated interfaces (for example, eth0 for ip data, eth1 for ip private cluster heartbeat) and these servers don't have a connection to SAN/NAS Storage so these servers just using their internal disks. Then our application team need to make their application active-standby that remain inside internal disk server (for example the mount point is /apps in internal disk). How to make this mount point sync to another node so that if node one went down, the node two will take the service of the application going up ? is that correct that we need to build red hat cluster under this circumstances ? what is the method of failover when one server down ?
if we need the licenses of high availability servers, what kind license subscription do we need ?
Hereby attachment picture of sample high availability topology.
![]()
Thanks in advance.
Regards,
Aldi
Responses
Hi Aldi,
Take a look at Red Hat storage.
Use setup two bricks for the application disk, one at each node and use the glusterfs volume to hold the application.
Still ask your self the question: "I want high availability, can I afford to loose an internal disk and then loose the application?" High available, sound to me Mission Critical.
Even Red Hat Storage is better put on external servers, these can be Virtual Machines.
Kind regards,
Jan Gerrit Kootstra
-edit-
Re-read question and realised GFS was out!
GlusterFS as mentioned above is a good option if no shared disk is available.
I assume iSCSI (or possibly even NFS) would be a viable option for the shared disk as well. If you are going to virtualize, then Red Hat Enterprise Virtualization is a great option which can use either NFS or iSCSI for the data store(s).
I agree with both responses. In particular Jan's question about what "high availability" actually means. Red Hat Storage (GlusterFS) can provide that resiliency (multiple copies across multiple physical systems, if that is your design) and remove that single-point-of-failure.
Jan/Pixel - do you know if RH Storage can do asynchronous remote data center copies?
Unfortunately due to lack of licenses I haven't had my hands on RH Storage.
I don't see why you couldn't use the replication/mirroring in GlusterFS to achieve asynchronous replication across data centers, the only caveat being low latency/high speed link between the two sites.
Gluster also supports Geo replication, but I believe it's only one way.
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
