[pve-devel] [PATCH qemu-server v3 1/6] fix #4225: qemuserver: drive: add parameter to mark drive required

[Daniel Herzig]

Fiona Ebner <f.ebner at proxmox.com> writes:

> The 'qemuserver' prefix in the commit title doesn't add any information
> and should not be there. Commit title prefixes are not for file names.
> This also doesn't fix the issue yet, so I'd also drop that prefix too.
>
Right, missed that redundancy. Will be fixed in some v4.

> Am 30.01.25 um 12:31 schrieb Daniel Herzig:
>> This commit add the parameter `essential` to mark a drive as required
>> for booting the VM.
>> 
>> Signed-off-by: Daniel Herzig <d.herzig at proxmox.com>
>> ---
>>  PVE/QemuServer/Drive.pm | 9 ++++++++-
>>  1 file changed, 8 insertions(+), 1 deletion(-)
>> 
>> diff --git a/PVE/QemuServer/Drive.pm b/PVE/QemuServer/Drive.pm
>> index 1041c1dd..38136787 100644
>> --- a/PVE/QemuServer/Drive.pm
>> +++ b/PVE/QemuServer/Drive.pm
>> @@ -266,7 +266,14 @@ my %drivedesc_base = (
>>  	verbose_description => "Mark this locally-managed volume as available on all nodes.\n\nWARNING: This option does not share the volume automatically, it assumes it is shared already!",
>>  	optional => 1,
>>  	default => 0,
>> -    }
>> +    },
>> +    essential => {
>> +	type => 'boolean',
>> +	description => 'Mark this iso volume as required for booting the VM.',
>
> Nit: Since the 'media=cdrom' option is used to decide this, I'd state
> "CD-ROM" here to be more precise. I'd also say "for starting the VM"
> rather than "for booting the VM". The device isn't necessarily involved
> into booting, but can still be considered essential to allow starting
> the VM.

Thanks for pointing this out, this part needs some re-working indeed.

>
>> +	verbose_description => 'If unset or set to 1, and the iso file is unavailable, the VM will not start.\nThis parameter is considered for cdrom iso drives only.',
>> +	optional => 1,
>> +	default => 1,
>> +    },
>>  );
>>  
>>  my %iothread_fmt = ( iothread => {
>
> There should be some checking/handling in the API endpoints and/or
> parse_drive() for making sure this can only be set in combination with
> 'media=cdrom'. There are already such checks in parse_drive() which will
> return undef in those cases, but please also issue a warning for a new
> such check so that it will be clear what went wrong ;)
>
I'll look this up and check against various ~qm set~ scenarios.

> Nit: Usually, it's nicer to have booleans be 0 if not present. So can we
> invert this e.g. "ignore-if-missing" (or "detach-if-missing" or
> "eject-if-missing")? Otherwise, maybe "essential-for-start" to be more
> descriptive?

Good idea, getting ready to invert my brains :)