[pve-devel] [RFC v6 qemu-server] VM protection mode

[Alen Grizonic]

On 09/10/2015 06:44 PM, Dietmar Maurer wrote:
>> @@ -889,6 +897,7 @@ my $update_vm_api  = sub {
>>   		$modified->{$opt} = 1;
>>   		$conf = PVE::QemuServer::load_config($vmid); # update/reload
>>   		if ($opt =~ m/^unused/) {
>> +		    &$check_protection($conf, "can't remove VM $vmid storage");
> strange error message, because you remove a 'volume' (or disk), not a 'storage'.
>

Yes, better.

>>   		    $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
>>   		    my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
>>   		    if (PVE::QemuServer::try_deallocate_drive($storecfg, $vmid, $conf,
>> $opt, $drive, $rpcenv, $authuser)) {
>> @@ -896,11 +905,14 @@ my $update_vm_api  = sub {
>>   			PVE::QemuServer::update_config_nolock($vmid, $conf, 1);
>>   		    }
>>   		} elsif (PVE::QemuServer::valid_drivename($opt)) {
>> +		    &$check_protection($conf, "can't remove VM $vmid storage");
> you remove a 'disk' (or 'drive'), but not a 'storage'.

Ok, I'll change also this one.

>
>>   		    $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
>>   		    PVE::QemuServer::vmconfig_register_unused_drive($storecfg, $vmid,
>> $conf, PVE::QemuServer::parse_drive($opt, $conf->{pending}->{$opt}))
>>   			if defined($conf->{pending}->{$opt});
>>   		    PVE::QemuServer::vmconfig_delete_pending_option($conf, $opt, $force);
>>   		    PVE::QemuServer::update_config_nolock($vmid, $conf, 1);
>> +		} elsif ($opt eq 'protection') {
>> +		    &$check_protection($conf, "can't delete VM $vmid protection option");
> Is this really helpful? Why do you think
>
> # qm set <vmid> --protection 0
>
> is better than
>
> # qm set <vmid> --delete protection
>
> IMHO both things are similar?

Both similar. I've put it here so they possibly use the

qm set <vmid> --protection 0

which is used also by the GUI, before they completely remove the option.

Feature to be discussed.



> Note: Some of our GUI widgets automatically send delete instead
> of setting the default value, so that can produce unexpected
> side effects.