Re: [pve-devel] [PATCH common] systemd: add helper to cleanup transient unit

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

Am 20/06/2023 um 17:00 schrieb Fiona Ebner:
> which combines the stop+wait logic previously present at the single
> call site of wait_for_unit_removed() in QemuServer.pm. It also does a
> reset-failed call first, to ensure a unit in a failed state is also
> cleaned up properly.
> 
> Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
> ---
>  src/PVE/Systemd.pm | 16 +++++++++++++++-
>  1 file changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/src/PVE/Systemd.pm b/src/PVE/Systemd.pm
> index 2517d31..327106f 100644
> --- a/src/PVE/Systemd.pm
> +++ b/src/PVE/Systemd.pm
> @@ -7,7 +7,7 @@ use Net::DBus qw(dbus_uint32 dbus_uint64 dbus_boolean);
>  use Net::DBus::Callback;
>  use Net::DBus::Reactor;
>  
> -use PVE::Tools qw(file_set_contents file_get_contents trim);
> +use PVE::Tools qw(file_set_contents file_get_contents run_command trim);
>  
>  sub escape_unit {
>      my ($val, $is_path) = @_;
> @@ -167,6 +167,20 @@ sub wait_for_unit_removed($;$) {
>      }, $timeout);
>  }
>  
> +sub cleanup_transient_unit($;$) {
> +    my ($unit, $timeout) = @_;
> +
> +    eval {
> +	my %param = ( outfunc => sub {}, errfunc => sub {} );
> +	# If the unit is in a failed state (e.g. after being OOM-killed), stopping is not enough.
> +	run_command(['/bin/systemctl', 'reset-failed', $unit], %param);
> +	run_command(['/bin/systemctl', 'stop', $unit], %param);
> +    };
> +
> +    # Issues with the above not being fully completed are rare, but not impossible, see bug #3733.
> +    wait_for_unit_removed($unit, $timeout);
> +}
> +
>  sub read_ini {
>      my ($filename) = @_;
>  

for the record, this the same in qemu-server directly for now, as talked off list,
didn't remembered that we got a run_command there already, still big thanks for
finding this!