Best practice for creating reliable mounts from SAN Storage - Redhat cluster

Latest response

I am working on a two-node Redhat cluster implementation which will host Oracle Databases. SAN team has presented LUNs to the two nodes and DBA is requesting the following.

Node1
/u101
/u102
/u103
/u104

TOTAL: Use the Pure Storage Volume provided by the storage team that is for Node1 at the requested size of 400 GB

Node2
/u201
/u202
/u203
/u204

TOTAL: Use the Pure Storage Volume provided by the storage team that is for Node2 at the requested size of 200 GB

My plan is to create a normal LVM setup on each box from the respective LUN and then create different mount points from the VG which can be managed as resources in Redhat Cluster. I have been reading on GFS2 and curious if that will be more reliable for hosting Oracle data as it will be a shared storage between two hosts?

Responses

Hello Amarpal Singh

If you are using SAN storage, likely it is a "fabric" where you ought to be using multipathing. These instructions are generic for setting up multipathing.

A few conditional things to keep in mind here:

  • If you are using a SAN and
  • If you are using a fabric (see link above)

If these are true, you ought to use the mulitipath method. You might have to identify the "world wide number" or WWN.

In short, this is what to consider with mounting SAN storage.

Regards

RJ

Hello,

Thanks for the response. I have configured the multipath already on both nodes and currently these mounts are being requested by the DBA team from the presented storage. I can see the LUNs presented to the hosts ( both nodes) and now getting ready to create the partitions so I can create PV/VG/LVMs and so I can mount them and later configure them as resources in Redhat Cluster so they can be managed via proper constraints. I was referred to GFS2 which is kind new to me so wondering if simple LVM setup is good enough instead of using GFS for our two-node redhat cluster setup.

Regards, AP

Hello Amarpal Singh,

I'd ask Red Hat directly (put in a case, or if you have a Technical Account Manager), or see if anyone else chimes in here. I know one thing you need with Gluster FS is to ensure you have it so that it is available if one specific node goes down. I helped set up some Gluster at one point, but I have minimal experience with Gluster. Sorry I missed your focus on Gluster in your original post.

Regards,

RJ

Amarpal Singh,

What version of Gluster are you using? I imagine you've probably read the Gluster documentation already...

Regards

RJ