Upgrading from version 3.4 to 7.2

[KoenDG]

So, I'm planning an upgrade from version 3.4 to 7.2

It would seem the best way to do this is by backing up all the VMs to a remote storage, shutting down the machine and then simply installing the most recent ISO, wiping all the old data.

And then restoring the backups.

I just want to make sure that would work.


Especially considering this:

https://pve.proxmox.com/wiki/Logical_Volume_Manager_(LVM)

For Proxmox VE versions up to 4.1, the installer creates a standard logical volume called “data”, which is mounted at /var/lib/vz.

Starting from version 4.2, the logical volume “data” is a LVM-thin pool, used to store block based guest images, and /var/lib/vz is simply a directory on the root file system.

On newer versions I'm seeing the output of `lsblk` showing a pve per VM under the /dev/mapper/data directory. In the old version, these are images under /var/lib/vz somewhere.

A potential issue​

The old machine has a mount that was added as extra storage. It's called /mnt/extrastorage and has an images folder, with the numbers of the VMs and their images under it. It is mounted from /etc/fstab

In particular: this storage is an on SSD. For faster I/O. And several VMs are specifically on that one, because of this fact.

I can back those up, but if I want to restore them, how will I add them to the data directory? Do I extend the data volume? If so, how? I saw it has 2 subgroups in lsblk, so that has me a bit confused.


So, just wanted to confirm this strategy of backup -> restore will work across such a big gap in versions. If snapshot is ok or it's better to stop them all and do a backup then.

And ask what to do about this mount.

 

[oguz]

hi,

just wanted to confirm this strategy of backup -> restore will work across such a big gap in versions.

yes it should work.

If snapshot is ok or it's better to stop them all and do a backup then.

depending on your (uptime) needs snapshot mode might be okay, but if you can put them offline and do a backup then it's probably safer.

A potential issue​

The old machine has a mount that was added as extra storage. It's called /mnt/extrastorage and has an images folder, with the numbers of the VMs and their images under it. It is mounted from /etc/fstab

In particular: this storage is an on SSD. For faster I/O. And several VMs are specifically on that one, because of this fact.

I can back those up, but if I want to restore them, how will I add them to the data directory? Do I extend the data volume? If so, how? I saw it has 2 subgroups in lsblk, so that has me a bit confused.

the old storage configuration should be mostly irrelevant for restoring the guests into your new installation (during a restore you can select which storage on your new node should the VM/CT be restored to).
so you probably won't have to extend volumes manually, most things should be already handled by our storage plugin.

one thing you could make sure would be if those disks in the extra storage are added correctly into the backup archives. if you find all your disks in the archives then generally you should be fine while restoring them.

if you have any questions don't hesitate to ask

 

[KoenDG]

the old storage configuration should be mostly irrelevant for restoring the guests into your new installation (during a restore you can select which storage on your new node should the VM/CT be restored to).
so you probably won't have to extend volumes manually, most things should be already handled by our storage plugin.

So, after initial setup, when first logging in, I add the mount to the storage.

Datacenter -> Storage -> Add

And give it "Disk Image" and "Container".

The restore command indeed has a dropdown

The SSD is currently nearly full, so I assume I'll have to wipe it and use it fresh as storage(after first backing up the VM images)?

Special situation​

One particular VM has 3 "disks" that are all on this SSD. These are used inside the VM as separate disks, as to partition things.

Looking at the backup with lzless, it shows the first disk, and then this:

scsi1:3r^C3.6p^C^G55G
socket<9D>#v<9C>^O5<89>^D28^T^A^L7G

The first disk is 72G, then the extra disks are 55G and 57G, so I guess this is them being mentioned.

However, for a total of 184GB, the backup is only 62GB. The log says it is 22% sparse.

I wonder if these disks were included.

The command from the logs was:

vzdump 120 --remove 0 --mode snapshot --compress lzo --storage backup_storage --node mynode

Considering the size, I get the impression that those 2 last disks are not included in the backup. Is there a way I could make sure?

one thing you could make sure would be if those disks in the extra storage are added correctly into the backup archives. if you find all your disks in the archives then generally you should be fine while restoring them.

All I see in the backup directory are log files and lzo files. And as stated before: looking at the lzo file with lzless shows a mention of the disks, but the low size is worrying.

EDIT: The cumulative size actually being used, inside the VM, is 96GB.

 

[oguz]

However, for a total of 184GB, the backup is only 62GB. The log says it is 22% sparse.

I wonder if these disks were included.

Considering the size, I get the impression that those 2 last disks are not included in the backup. Is there a way I could make sure?

you can try restoring it to be sure

And as stated before: looking at the lzo file with lzless shows a mention of the disks, but the low size is worrying.

keep in mind that the backup archive is also compressed.