R7x / R8x Installation Differences

[Daniel_Collins]

This post is more about some of my findings, rather than directing a question to anyone. Please correct me where I am wrong, as this is only my observations.

While watching an R80.20 upgrade occur on an R77.30 box and thought it just might be of interest to people.

I originally moaned at the fact that R80.20 wanted to take a snapshot of the system, before attempting to upgrade (because I'd historically stollen free space from lvm to give to /var/log/ in VMware based systems, as most customers use VM snapshots or something like Veeam to backup).

It turns out, it needs it to do the uprade. Traditionally, the CP upgrades were proper RHEL upgrades in that it had a bunch of RPM's and scripts that It simply upgraded packages on the system with. When upgrading to R80.20 from say the R77.30 box I am looking at it:

- Creates a new LVM partition (In my case, mounted as /mnt/fcd)

- Copies the installation to it and all local system confiugration files

- Exports the database in R80.20 format

- Saves all of this to the new LVM partition

- Reboots to the new partition

- Completes a few tidy up bits

- Re-imports the database into the new partition

- Starts up as normal

- Removes the old partition, renames the volume group to vg_splat

Sneaky..

[PhoneBoy]

The good thing about this approach is that reverting an upgrade is pretty straightforward: just revert the snapshot.

With an in-place upgrade, reverting is significantly more complex.

[JozkoMrkvicka]

In case of Major Upgrade (R77.30 -> R80.20), there is always snapshot created which can be reverted back in case something went wrong during the upgrade.

[Daniel_Collins]

It's great and leaves a nice clean break for roll-back without too much concern that might affect the pre-upgraded system.

It's almost like the system that F5 use in image management.