Dear All,
Looks like I can do a snapshot on LVM (not thin)
below are the details
Am I missing something?
Looks like I can do a snapshot on LVM (not thin)
below are the details
Am I missing something?
Last edited:
Yes, exactly, i use raw and qcow2 vm disks on ext4 over lvm. Can you please point me to the right direction where I can get more info about that? You state it is as it's an expected behavior, but it is a surprise for me.If you can use snapshot, you either use lvm thin, or file based images with qcow2, on a filesystem, on an lvm volume. But you can't use snapshot with non thin lvm directly
It's absolutely clear now.If you use file based images, it doesn't really matters what's under the filesystem (could be a standard lvm, thin lvm, raw partition, nfs etc...). You just rely on the qcow2 format to provide snapshots features. But proxmox can also use raw lvm volume as images (in which case there's no filesystem, so no file either). In this case, standard lvm can't use snapshots (but has the advantage of being shareable, for example on a shared iscsi lun). And thin lvm can use snapshots (lvm level snapshots)
In this case, standard lvm can't use snapshots (but has the advantage of being shareable, for example on a shared iscsi lun). And thin lvm can use snapshots (lvm level snapshots)
Hm. You mean snapshot on thick LVM is not a point in time logical snapshot, but more a full backup of vm disk + ram?You can also create snapshots on thick LVM, but it is as with thin LVM not cluster aware and will therefore not work in a clustered setup. LVM thick snapshots are different than "normal" VM snapshots and do have a size. If the snapshot is full, it is not working anymore. Totally incompatible with the modern age of immutable VM snapshots.
Hm. You mean snapshot on thick LVM is not a point in time logical snapshot, but more a full backup of vm disk + ram?
No, unfortunately neither. A snapshot is the place where all changed blocks will be written to and if the space is exhausted, your snapshot is useless.
LVM thick snapshots basically only serve on purpose - getting a frozen view of the LV so that you can make a backup. PVE uses other mechanisms for that (Qemu copy-before-write), They are not workable for snapshots that are kept around, so they are not supported. Of course you can manually create such a snapshot on the storage layer, PVE will not interfere with it in any way AFAIK. I am not sure how it might negatively affected migration in case the VG is shared across a cluster via iSCSI.Besides that, is there any reason that Proxmox doesn't supports thick LVM snapshots, esspecially in a LVM-over-iSCSI scenario? Is it safe to create them manually via the LVM tools?
thick LVM snapshots are also something completely different than thin snapshots. They have a given space and - as fabian already wrote - are only there to get a frozen view. If you write a block on your logical volume, the old block will be written into the snapshot, so that it'll always represent the data of the snapshot but if you run out of space for changes, your snapshot will get invalid and useless.LVM thick snapshots basically only serve on purpose - getting a frozen view of the LV so that you can make a backup. PVE uses other mechanisms for that (Qemu copy-before-write), They are not workable for snapshots that are kept around, so they are not supported. Of course you can manually create such a snapshot on the storage layer, PVE will not interfere with it in any way AFAIK. I am not sure how it might negatively affected migration in case the VG is shared across a cluster via iSCSI.