Re: [pve-devel] [PATCH qemu-server 3/3] api: include not mapped resources for running vms in migrate preconditions

[Fiona Ebner <f.ebner@proxmox.com>]

Am 20.03.24 um 13:51 schrieb Dominik Csapak:
> so that we can show a proper warning in the migrate dialog and check it
> in the bulk migrate precondition check
> 
> the unavailable_storages and allowed_nodes should be the same as before
> 
> Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
> ---
> not super happy with this partial approach, we probably should just
> always return the 'allowed_nodes' and 'not_allowed_nodes' and change
> the gui to handle the running vs not running state?

So not_allowed_nodes can already be returned in both states after this
patch. But allowed nodes still only if not running. I mean, there could
be API users that break if we'd always return allowed_nodes, but it
doesn't sound unreasonable for me to do so. Might even be an opportunity
to structure the code in a bit more straightforward manner.

> 
>  PVE/API2/Qemu.pm | 27 +++++++++++++++------------
>  1 file changed, 15 insertions(+), 12 deletions(-)
> 
> diff --git a/PVE/API2/Qemu.pm b/PVE/API2/Qemu.pm
> index 8581a529..b0f155f7 100644
> --- a/PVE/API2/Qemu.pm
> +++ b/PVE/API2/Qemu.pm
> @@ -4439,7 +4439,7 @@ __PACKAGE__->register_method({
>  	    not_allowed_nodes => {
>  		type => 'object',
>  		optional => 1,
> -		description => "List not allowed nodes with additional informations, only passed if VM is offline"
> +		description => "List not allowed nodes with additional informations",
>  	    },
>  	    local_disks => {
>  		type => 'array',
> @@ -4496,25 +4496,28 @@ __PACKAGE__->register_method({
>  
>  	# if vm is not running, return target nodes where local storage/mapped devices are available
>  	# for offline migration
> +	my $checked_nodes = {};
> +	my $allowed_nodes = [];
>  	if (!$res->{running}) {
> -	    $res->{allowed_nodes} = [];
> -	    my $checked_nodes = PVE::QemuServer::check_local_storage_availability($vmconf, $storecfg);
> +	    $checked_nodes = PVE::QemuServer::check_local_storage_availability($vmconf, $storecfg);
>  	    delete $checked_nodes->{$localnode};
> +	}
>  
> -	    foreach my $node (keys %$checked_nodes) {
> -		my $missing_mappings = $missing_mappings_by_node->{$node};
> -		if (scalar($missing_mappings->@*)) {
> -		    $checked_nodes->{$node}->{'unavailable-resources'} = $missing_mappings;
> -		    next;
> -		}
> +	foreach my $node ((keys $checked_nodes->%*, keys $missing_mappings_by_node->%*)) {

Style nit: please use 'for' instead of 'foreach'

Like this you might iterate over certain nodes twice and then push them
onto the allowed_nodes array twice.

> +	    my $missing_mappings = $missing_mappings_by_node->{$node};
> +	    if (scalar($missing_mappings->@*)) {
> +		$checked_nodes->{$node}->{'unavailable-resources'} = $missing_mappings;
> +		next;
> +	    }
>  
> +	    if (!$res->{running}) {
>  		if (!defined($checked_nodes->{$node}->{unavailable_storages})) {
> -		    push @{$res->{allowed_nodes}}, $node;
> +		    push $allowed_nodes->@*, $node;
>  		}
> -
>  	    }
> -	    $res->{not_allowed_nodes} = $checked_nodes;
>  	}
> +	$res->{not_allowed_nodes} = $checked_nodes if scalar(keys($checked_nodes->%*)) || !$res->{running};

Why not return the empty hash if running? The whole post-if is just
covering that single special case.

> +	$res->{allowed_nodes} = $allowed_nodes if scalar($allowed_nodes->@*) || !$res->{running};

Nit: Right now, $allowed_nodes can only be non-empty if
!$res->{running}, so the first part of the check is redundant.

>  
>  	my $local_disks = &$check_vm_disks_local($storecfg, $vmconf, $vmid);
>  	$res->{local_disks} = [ values %$local_disks ];;