Hello,

I need specific advise from a Linux Architect / Sr. Admin taking part in designing INF Arch for the scenario mentioned here. Please read carefully and answer questions.

Scenario: The project will be implemented in a VMware environment. We have a project requirement to build Linux servers to support a vendor supported application to run on RHEL 6.x. We will need to build 2 Linux nodes and each individual node needs to have access to a shared storage location (5 GB) where data will be maintained and needs to be read and written at the same time by both the nodes. Note, we don't need Linux clustering. So, both nodes will be Active independent nodes at all times. Any nodes should be able to read and write to this shared storage anytime. The shared storage should be accessible for both reads and writes at all times even if one of the Linux nodes goes down/offline. Also, we will need to replicate this shared storage to another Data Center if we want to bring up the DR Linux nodes in a DR situation. Application has file locking capabilities. (To be checked with vendor, not confirmed yet).

Questions:

Q1. I believe we will NOT need Linux Clustering as both Linux nodes are independent and are not in a Active and Passive configuration but capable of reading and writing to the shared storage area. Correct?

Q2. I know the advantages and disadvantages of GFS2 over NFS. However, to keep things simple, can't I use just NFS versus GFS2?

Q3. Should I use GFS2 as a mandate for file locking?

Q4. Does NFS not have file locking for 2 independent nodes with shared storage along with reads and writes at the same time?

===================================================================

I'm looking for some deep insights as this pertains to Architecture Design.

I understand the following:

NFS is a shared file system like SMB or CIFS. All clients who mount the same share will be able to see the same data. NFS uses a locking mechanism to determine who is writing to files so that more than one client cannot write to the same file at the same time. Multiple clients can read the same file at the same time though. GFS2 is block level vs file level but similar functionality where multiple clients can read and write to the same block device and there is a locking mechanism in GFS2 that prevents to clients from writing to the same file at the same time.

===================================================================

Q6. If this is the case, NFS should suffice to my requirement. Correct?

Q7. Why should I care about GFS2 in this scenario? Is it just because it is block level and is more reliable and provides better performance and scale well? Possibly, I don't need (or) may be it is needed? How to determine? I have worked with NFS earlier and had not had severe performance issues. Does NFS fit this scenario?

Please provide insights. Thanks for your inputs in advance!

The answer depends if you have central storage with NFS service (NetApp ?)
If you have one, then answer is NFS + replica to DR site builtin.

If you do not have reliable NFS server, then use GFS2. This, however, require clusterware installed and configured, because GFS2 use it's infrastructure.
When you talking about VM, it is not so easy to make shared LUN between VMs.
Usually it is about not use vmotion and both VMs will running on same host, that is bad.
Then option DRBD comes to mind.

I'd recommend you use NFS if you have reliable NFS server.
If you will go with GFS2+DRBD, read my POC: Building active-active RedHat 6 Cluster with GFS2 over DRBD

1 members found this post helpful.