UNIX for Advanced & Expert Users

All,

We have a physical server (Lenovo 3950) that we need to upgrade the OS from RHEL 6 to RHEL 7. We do not want to do an in-place upgrade but rather start with a fresh OS install. The plan is to remove the two local drives (Raid 1) and install two new drives for the fresh RHEL 7 install. The idea being we can put the original drives back in the server if anything goes wrong. This part is pretty straight forward, but our question concerns the presentation of the LUNs that are used for the database (SAP HANA) on the old OS.

After the installation of the OS we will want to present snapshots of the original LUNs to the new RHEL 7 system. The LUNs reside on an EMC Xtreme IO array. The LUNs have LVM VGs and LVs with XFS filesystems. We would like to understand what we should expect with the LUNs presented to the new OS.

- Would this process be as simple as using the same /etc/fstab line from the old OS in the new OS and it recognize the LVM configuration and data as before? We are defining the mount via UUID on the current RHEL 6 OS.

- Do we need to or would it be better practice to do some type of lvconvert (mirror) or similar on the new OS? I recall doing a "mklvcopy" for this type of need in AIX, similar in RHEL 7?

We are going to test this in a sandbox environment but any guidance or any advise from experience would be greatly appreciated.

Thanks in advance,

HB

With the costs of disks so cheap these days, you are better to clone the disks and put them in a development / test server instead of removing the critical disks and testing them, with a plan to rush them back to the original server if something goes wrong.

This is especially true if you are talking about a critical production asset.

If you are talking about a "mom and pop" server which does not serve a critical business need, then you can certainly consider some "quick, disk shuffle" method, but not for a business-critical production server.

Just fully recreate the current computer with clones of the application and data disks in a your new OS build / server; and then cut over too it when all is tested and ready to go live (syncing data as necessary).

Thank you for the response.

We were not planning on testing the removal of the disks and installing a the new OS. That part is not a concern. We will test the presentation of the snapped LUNs to a new server/OS.

Our unknown is the presentation of the LUNs and how RHEL 7 will 'see' the LVM and data. We want to have all of the steps defined for when we are ready to do the process in the critical environment..

HB

Presenting snapshot luns will work fine, as long as you don't present source and snapshots to the same box.

Rescan the fabric, followed by pvscan, then vgimport <your_group_from_pvscan>
Take care of the /etc/fstab with a help of blkid to list the UUIDs od the logical volumes.

Hope that helps
Regards
Peasant.

Thank you @Peasant for the reply. I wanted to provide an update...

The process was actually much easier than expected. We just needed to:

1. Snapshot the original LUN.
2. Create the same mount point on the new server.
3. Copy the /etc/fstab entry from original server and write it to the /etc/fstab on the new server.
4. Present the snapshot LUN to the new server.
5. Reboot the new server.

This allowed us to to see the original data as expected.

Thanks again for the help..

HB