Elevate CentOS 7.x to AlmaLinux 8.0 at DigitalOcean issues

eugenevdm.host

• June 18, 2023 05:21

I'm sharing my experience elevating to AlmaLinux at DigitalOcean. The first lesson I learnt was rather do a snapshot before you do anything. Unfortunately to be consistent with the advice given by DigitalOcean, one has to shut down the droplet. Our 60 GB droplet with 40 GB of used space takes almost an hour to take a snapshot. Already here I have 45 minutes on downtime which is an issue for my always on clients. Next please be aware of the following, according to DigitalOcean support: [QUOTE]An in-place upgrade of your CentOS distribution to AlmaLinux is also not something we recommend or can support. Instead, we would recommend that you deploy a new AlmaLinux Droplet and migrate your data.
So what this means is if you nuke your WHM installation, don't ask DigitalOcean for help. Next are lessons are per server. So far I have had one success with fiddling, and another fail. In total I've done four elevations of which three succeeded, however, two of those weren't with DigitalOcean. The two that weren't with DigitalOcean were with other hosts, a Xen host and a KVM host. In both instances it succeeded out of the box, but both servers were pretty simple DNS Only servers. Server 1 Issues

• Didn't reboot after stage 3. Stuck on dismounting volume. Force switch off at DigitalOcean control panel, whereupon it continued. Glad that I could actually SSH to the server again.
• After this force reboot, elevate complained that stage 4 didn't work. What became quickly clear is that AlmalLinux was now try to use DNF commands, but DNF was broken. Actually even trying to use YUM it complained about the same problem. Thankfully I tried using CURL and then saw CURL was also broken with the same problem. This was a clue something fundamental, in the case a library, was broken.

Basically what I learnt here is `/usr/bin/dnf clean all` fails but basically some libraries are missing. The actual error was: [QUOTE]ImportError: libpsl.so.0: cannot open shared object file: No such file or directory
At first I went into a mild panic and logged a ticket with cPanel, but also started troubleshooting. This article linked to an old ticket of us was a major breakthrough: Update fails, libcurl / libpsl issues ...but had to bring the server up again due to prolonged downtime. To understand if I had a once off problem, I created and restored the snapshot and tried again. Again the system failed at exactly the same point: [QUOTE]XFS (vda1): WARNING: Reset corrupted AGFL on AG 7. 6 blocks leaked. Please unmount and run xfs_repair.
However, this time around it didn't boot anymore after trying to init `/usr/sbin/init ` Lessons Learnt - DigitalOcean doesn't support or recommend elevate - Snapshots are a life saver - Get ready for lots of downtime when doing AlmalLinux elevate - Probably have to go the backup and restore route I'm not super happy about the corrupted 'AGFL' stuff, so let's see if DigitalOcean is willing to help me fix it.

• 

  cPRex Jurassic Moderator

  • June 19, 2023 15:11

  Hey there! We do have an option for server providers to create a touch file as outlined in our documentation at the following link which will keep users from running ELevate if they did want to restrict that:

  0
• 

  otakudes

  • June 27, 2023 01:41

  Thanks for posting. I had the same issues attempting to update to Rocky and Alma. I forced rebooted from DO's admin panel and could log back in with SSH but no access to cpanel on the web.

  0

Please sign in to leave a comment.