Re: [PATCH v3 pve-qemu qemu-server 0/6] bypass host page cache for restore on zvol

[Thomas Lamprecht <t.lamprecht@proxmox.com>]

On 20/08/2026 11:56, Lukas Sichert wrote:
> On 2026-08-20 11:05, Christian Ebner <c.ebner@proxmox.com> wrote:
> 
>> This patch series skips writing to the host page cache for restores
>> to zvols, with the intention to avoid potential I/O delay issues for
>> other ZFS-backed VMs during page writeback on certain setups, as
>> reported in enterprise support and reproduced internally.
>>
>> This patch series adds an optional no-cache parameter for pbs-restore
>> command invocation, setting the BDRV_O_NOCACHE flag for the block
>> device being restored to and sets `cache=none` for vma restores.
> 
> When I test patched debs on my testmachines I normally bump the version
> with a local<x> suffigx. package_name_1.2.3 is deployed as
> package_name_1.2.3+local1. When starting a restore with this and v2, I got
> the error 'TASK ERROR: internal error: cannot check version of invalid
> string 'unknown' at /usr/share/perl5/PVE/QemuServer/Helpers.pm line 104.'
> 
> Christian proposed the following diff off-list, which worked for me.
> 
> diff --git a/src/PVE/QemuServer/Helpers.pm b/src/PVE/QemuServer/Helpers.pm
> index 5065a457..1ed48a3b 100644
> --- a/src/PVE/QemuServer/Helpers.pm
> +++ b/src/PVE/QemuServer/Helpers.pm
> @@ -68,7 +68,7 @@ sub kvm_user_version {
>      $kvm_mtime->{$binary} = $st->mtime;
> 
>      my $binary_version_re = qr/(\d+\.\d+(\.\d+)?)(\.\d+)?/;
> -    my $package_version_re = qr/\(pve-qemu-kvm_(\d+\.\d+(\.\d+)?-\d+)\)/;
> +    my $package_version_re = qr/\(pve-qemu-kvm_(\d+\.\d+(\.\d+)?-\d+)(\+[a-zA-Z0-9]+)?\)/;
> 
>      my $code = sub {
>          my $line = shift;
> 
> I wouldn't consider this a blocker, but it would be a nice to have in a
> potential next version.


meh, that's far from great - re-implementing all off debians
package version seems also futile and hard to maintain; so either
we use "dpkg --compare-versions ..." directly via run_command or
just bump the versioned dependency in d/control here - even though
the latter makes downgrading slightly more work, it is IMO still more
preferable for this here, and it would allow us to drop like three
patches FWICT.