[pve-devel] [PATCH qemu-server v3] fix #2788: do not resume vms after backup if they were paused before

[Fabian Ebner]

One minor thing inline, and that likely needs a fix elsewhere if we even 
want to bother. Other than that:

Tested-by: Fabian Ebner <f.ebner at proxmox.com>

On 20.01.21 13:32, Dominik Csapak wrote:
> by checking if the vm is paused at the beginning and skipping the resume
> now we also skip the qga freeze/thaw (which cannot work if the vm is
> paused)
> 
> moved the 'vm_is_paused' sub from the api to PVE/QemuServer.pm so it is
> available everywhere we need it.
> 
> since a suspend backup would pause the vm anyway, we can skip that step
> also
> 
> Signed-off-by: Dominik Csapak <d.csapak at proxmox.com>
> ---
> changes to v2:
> * actually include code changes (git add is not optional), sry for the noise
> 
>   PVE/API2/Qemu.pm         | 14 ++------------
>   PVE/QemuServer.pm        | 10 ++++++++++
>   PVE/VZDump/QemuServer.pm | 11 +++++++++--
>   3 files changed, 21 insertions(+), 14 deletions(-)
> 
> diff --git a/PVE/API2/Qemu.pm b/PVE/API2/Qemu.pm
> index e2d2d67..623c998 100644
> --- a/PVE/API2/Qemu.pm
> +++ b/PVE/API2/Qemu.pm
> @@ -2402,16 +2402,6 @@ __PACKAGE__->register_method({
>   	return $rpcenv->fork_worker('qmreset', $vmid, $authuser, $realcmd);
>       }});
>   
> -my sub vm_is_paused {
> -    my ($vmid) = @_;
> -    my $qmpstatus = eval {
> -	PVE::QemuConfig::assert_config_exists_on_node($vmid);
> -	mon_cmd($vmid, "query-status");
> -    };
> -    warn "$@\n" if $@;
> -    return $qmpstatus && $qmpstatus->{status} eq "paused";
> -}
> -
>   __PACKAGE__->register_method({
>       name => 'vm_shutdown',
>       path => '{vmid}/status/shutdown',
> @@ -2480,7 +2470,7 @@ __PACKAGE__->register_method({
>   	#
>   	# checking the qmp status here to get feedback to the gui/cli/api
>   	# and the status query should not take too long
> -	if (vm_is_paused($vmid)) {
> +	if (PVE::QemuServer::vm_is_paused($vmid)) {
>   	    if ($param->{forceStop}) {
>   		warn "VM is paused - stop instead of shutdown\n";
>   		$shutdown = 0;
> @@ -2556,7 +2546,7 @@ __PACKAGE__->register_method({
>   	my $node = extract_param($param, 'node');
>   	my $vmid = extract_param($param, 'vmid');
>   
> -	die "VM is paused - cannot shutdown\n" if vm_is_paused($vmid);
> +	die "VM is paused - cannot shutdown\n" if PVE::QemuServer::vm_is_paused($vmid);
>   
>   	die "VM $vmid not running\n" if !PVE::QemuServer::check_running($vmid);
>   
> diff --git a/PVE/QemuServer.pm b/PVE/QemuServer.pm
> index 54278e5..d7ee05f 100644
> --- a/PVE/QemuServer.pm
> +++ b/PVE/QemuServer.pm
> @@ -7382,4 +7382,14 @@ sub complete_migration_storage {
>       return $res;
>   }
>   
> +sub vm_is_paused {
> +    my ($vmid) = @_;
> +    my $qmpstatus = eval {
> +	PVE::QemuConfig::assert_config_exists_on_node($vmid);
> +	mon_cmd($vmid, "query-status");
> +    };
> +    warn "$@\n" if $@;
> +    return $qmpstatus && $qmpstatus->{status} eq "paused";
> +}
> +
>   1;
> diff --git a/PVE/VZDump/QemuServer.pm b/PVE/VZDump/QemuServer.pm
> index b322701..b5e74d3 100644
> --- a/PVE/VZDump/QemuServer.pm
> +++ b/PVE/VZDump/QemuServer.pm
> @@ -62,8 +62,11 @@ sub prepare {
>   	if defined($conf->{name});
>   
>       $self->{vm_was_running} = 1;
> +    $self->{vm_was_paused} = 0;
>       if (!PVE::QemuServer::check_running($vmid)) {
>   	$self->{vm_was_running} = 0;
> +    } elsif (PVE::QemuServer::vm_is_paused($vmid)) {
> +	$self->{vm_was_paused} = 1;
>       }
>   
>       $task->{hostname} = $conf->{name};
> @@ -176,12 +179,16 @@ sub start_vm {
>   sub suspend_vm {
>       my ($self, $task, $vmid) = @_;
>   
> +    return if $self->{vm_was_paused};
> +
>       $self->cmd ("qm suspend $vmid --skiplock");
>   }
>   
>   sub resume_vm {
>       my ($self, $task, $vmid) = @_;
>   
> +    return if $self->{vm_was_paused};
> +
>       $self->cmd ("qm resume $vmid --skiplock");
>   }
>   
> @@ -794,7 +801,7 @@ sub _get_task_devlist {
>   
>   sub qga_fs_freeze {
>       my ($self, $task, $vmid) = @_;
> -    return if !$self->{vmlist}->{$vmid}->{agent} || $task->{mode} eq 'stop' || !$self->{vm_was_running};
> +    return if !$self->{vmlist}->{$vmid}->{agent} || $task->{mode} eq 'stop' || !$self->{vm_was_running} || $self->{vm_was_paused};

It's possible to suspend and resume VMs during backup. E.g. somebody 
else can resume the VM (even without using --skiplock) right after the 
backup task called suspend_vm. Relying on vm_was_paused here leads to 
skipping the fs-freeze for such an (admittedly unlikely) edge case.

>   
>       if (!PVE::QemuServer::qga_check_running($vmid, 1)) {
>   	$self->loginfo("skipping guest-agent 'fs-freeze', agent configured but not running?");
> @@ -872,7 +879,7 @@ sub register_qmeventd_handle {
>   sub resume_vm_after_job_start {
>       my ($self, $task, $vmid) = @_;
>   
> -    return if !$self->{vm_was_running};
> +    return if !$self->{vm_was_running} || $self->{vm_was_paused};
>   
>       if (my $stoptime = $task->{vmstoptime}) {
>   	my $delay = time() - $task->{vmstoptime};
>