Install specific proxmox/ceph versions

[bagels1b]

I have a virtual cluster within my active ProxMox cluster. The versions they are running have gotten out of sync and I want to get them back in sync before I upgrade my live cluster to ProxMox v8.3.1 and Ceph Squid so I can do a test upgrade in my virtual cluster. I need my virtual cluster to go from 8.2.2 to 8.2.4 and Ceph to actually downgrade from 18.2.4 to 18.2.2. I assume if I just do apt full-upgrade it will take it to 8.3.1. Is that correct? Is it possible to downgrade Ceph Reef?

Thanks

 

[dakralex]

Hi!

I assume if I just do apt full-upgrade it will take it to 8.3.1

Yes, depending on the apt-cache (the last time when apt update has been called), it will upgrade the packages to the newest known versions.

Is it possible to downgrade Ceph Reef?

I see no reason why not, especially for minor versions.

I'm still puzzled why you want to sync the versions to patch version, but apt/apt-get allows to provide version numbers with a modifier, e.g. apt-get install pve-manager=8.2.10 for the last 8.2.* version of the pve-manager package.

 

[bagels1b]

Thanks @dakralex. I did apt-get install pve-manager=8.2.4 so my test virtual cluster now matches my production cluster pve version but my kernel's don't match. My production cluster version is 6.8.8-2 while the test cluster is 6.8.4-2. If I do 'apt install pve-kernel-6.8' it wants to upgrade me to 6.8.12-8. If I do 'apt install pve-kernel-6.8.8-2' it can't find that package. Is there a way to get my kernel to 6.8.8-2?

 

[dakralex]

Is there a way to get my kernel to 6.8.8-2?

There are multiple packages related to the Proxmox VE kernel:

• proxmox-default-kernel depends on the package with the current default version (which currently is 6.8 as of PVE 8.2 and 8.3).
• proxmox-kernel-X.Y depends on the package with the current latest version in the series (i.e. proxmox-kernel-6.8 depends on proxmox-kernel-6.8.12-8).
• proxmox-kernel-X.Y.Z-R-pve is the actual package with the specific kernel version for non-secure boot installations.
• proxmox-kernel-X.Y.Z-R-pve-signed is the actual package with the specific kernel version for secure boot installations.

Depending on your setup (whether you have setup Secure Boot in your test cluster), you just need to install the package with the specific kernel version, i.e. apt install proxmox-kernel-6.8.8-2-pve or apt install proxmox-kernel-6.8.8-2-pve-signed.

 

[bagels1b]

Do you know what the procedure is to downgrade Ceph? I've searched around but most everything is related to upgrading reef to squid.

 

[dakralex]

For minor and patch versions (e.g. 18.2.4 to 18.2.2), downgrading should be fine with the exact same procedure as with the pve-manager package. I've never heard of downgrading major versions (e.g. squid to reef), nor do I believe that anyone bothered implementing that.

 

[Maximiliano]

Hello, please always upgrade package when you need to get things in sync. Downgrading packages can have unintended consequences in rare scenarios. I would suggest to upgrade everything to the latest version via

apt update && apt full-upgrade

to get everything in sync.

 

[bagels1b]

Hi @Maximiliano. My intention with the downgrade was to get my test cluster in sync with my production cluster so I could test the upgrade before doing it on the production cluster. I intend to upgrade to the latest but I was hoping to test it first.

Regarding "apt update && apt full-upgrade". I also see "apt dist-upgrade" in a lot of places. Is it the same thing?

 

[dakralex]

Regarding "apt update && apt full-upgrade". I also see "apt dist-upgrade" in a lot of places. Is it the same thing?

apt full-upgrade is the same thing as apt dist-upgrade. The latter is there for muscle memory, because of apt-get dist-upgrade.