Hi everyone,
This is going to be my first post on this forum and I apologize beforehand because I feel it is going to be a long one.
I’ve just been made responsible for a company’s email server and dns running MIAB v57a running on a Ubuntu 18.04.6 LTS virtual machine hosted on Linode.
I want to upgrade it to the latest version: v71a. I’ve read through the documentation on how to perform the upgrade, but here is where it gets tricky.
This server has 102 domains an a whole lot of email users and aliases. It works great, with little or no problems at all. The big issue here is our storage and the integrated backups MIAB has. We have over 7TB of data in emails. My instructions have been to preserve as much history as it is possible, and we have no issue with having to pay for all that storage. This storage is an attached volume to the VM and we resize it whenever it ist needed with no issues.
The problem is that the guide for upgrade states that we have to perform a backup with MIAB’s built in tool. The issue here is that performing the backup this way will take for ever and it is going to take a lot downtime, which we can´t afford to have.
My question is this: is it possible to just unmount the volume where we have the user-data directory, perfrom the backup with MIAB’s built-in tool and then follow all the remaining steps on the upgrade guide, and then mount the volume on the new box for MIAB with the new version and OS?
Thank you beforehand and I hope someone can help me.
Cheers.
I’m almost afraid to answer with that much mail on the line
My take is that indeed you can unmount the volume, and mount it on the new VM. Then the installation needs to be performed with the mounted volume again (it will update configs and databases where necessary)
You are going to do a dry run before doing this for real, right? RIGHT?
Thanks for your reply. Yes. This is what I was planning on doing:
For the test run, I’ve already deployed a new VM with the same specs, only difference is the OS (Ubuntu 22.04)
Sync the whole storage volume to a backup one on the same cloud.
Perform the backup (without the email data)
Restore the backup on the new VM
Mount the backup volume
Test if it works. (No idea how to do this, since I really dont want to move the public IP of the production box to the new box until I’m absolutely sure that it works).
Does this make sense? My gut tells me that this should go without any problems, unless MIAB’s backup process somehow needs to see the whole directory tree inside user-data.
No. We are not using NextCloud, so one less thing to worry about.
I’ve noticed that when I unmount the volume that has the user-data directory, I have the user-data directory still there with the default values. I was thinking on copying all files and directories to this original user-data directory, excepting all email files, that way, the backup operation shouldn’t take as long.