PVE 2.2 restore bug

[rengiared]

hi,

it seems like there is a restore bug in 2.2
i have a cluster of 3 nodes and backuped the system 210 on node 2
after the backup was successfull i did a restore on node 3 with the id 303
the restore itself was ok but the machine couldn't be started cause in the config file the wrong hdd entry was set

in the restored 303 config file the entry for the hdd was like this: 210/vm-210-disk-1.qcow2

no big deal, just a quick change from 210 to 303 and the problem was solved, but i think this is a new bug in 2.2 cause that happened never again before

to be honest, i havent had the time to try it a second time, but will do it this evening


hth and tia

 

[dietmar]

in the restored 303 config file the entry for the hdd was like this: 210/vm-210-disk-1.qcow2

We do not add entries like that. Maybe you manually edited that files? Please can you post the whole VM config file?

 

[rengiared]

hello,

no, i haven't done any manual entries in these config file cause the vm is only 2 days old

bootdisk: sata0
cores: 2
ide2: pve:iso/en_windows_server_2012_x64_dvd_915478.iso,media=cdrom,size=3608574K
memory: 4096
name: dc03
net0: e1000=62:DB:64:0D:FF:06,bridge=vmbr0
onboot: 1
ostype: win8
sata0: local:303/vm-303-disk-1.qcow2,size=50G
sockets: 1
startup: order=4

before i changed it to 303 the line with the hdd looked like this
sata0: local:210/vm-210-disk-1.qcow2,size=50G

in the meantime i did another restore of the backup on node 1 and the same problem appeared again
so for me its comprehensible

this is what it looks like after the restore in the gui

 

[dietmar]

OK, found the bug. This only happens for 'SATA' drives.

I uploaded a fix to the pvetest repository. We will try to move that into stable next week.

 

[rengiared]

fine, thank you
one additional question: is it by design that now qcow2s are as big as they are defined?
in the past they grew in size only when necessary?

 

[dietmar]

one additional question: is it by design that now qcow2s are as big as they are defined?
in the past they grew in size only when necessary?

Yes, we pre-allocate metadata now. But file is still small because it uses unix 'sparse file' feature ("check with 'du -sh filename')

 

[rengiared]

perfect, thank you for you explanation