From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gate001.proxmox.com (gate001.proxmox.com [IPv6:2a0f:8001:1:32::40]) by lore.proxmox.com (Postfix) with ESMTPS id B6B991FF0B2 for ; Tue, 22 Sep 2026 10:44:12 +0200 (CEST) Received: from gate001.proxmox.com (localhost.localdomain [127.0.0.1]) by gate001.proxmox.com (Proxmox) with ESMTP id 02E79214F5; Tue, 22 Sep 2026 10:44:10 +0200 (CEST) Mime-Version: 1.0 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=UTF-8 Date: Tue, 22 Sep 2026 10:44:04 +0200 Message-Id: Subject: Re: [PATCH qemu-server 1/1] fix #8030: snapshot: avoid orphaned vmstate volume on failure From: "Elias Huhsovitz" To: =?utf-8?q?Michael_K=C3=B6ppl?= , X-Mailer: aerc 0.20.0 References: <20260914140547.615346-1-m.koeppl@proxmox.com> In-Reply-To: <20260914140547.615346-1-m.koeppl@proxmox.com> X-Bm-Milter-Handled: 55990f41-d878-4baa-be0a-ee34c49e34d2 X-Bm-Transport-Timestamp: 1790066644533 X-SPAM-LEVEL: Spam detection results: 0 AWL 0.613 Adjusted score from AWL reputation of From: address DMARC_MISSING 0.1 Missing DMARC policy KAM_DMARC_STATUS 0.01 Test Rule for DKIM or SPF Failure with Strict Alignment (newer systems) RCVD_IN_DNSWL_MED -2.3 Sender listed at https://www.dnswl.org/, medium trust SPF_HELO_NONE 0.001 SPF: HELO does not publish an SPF Record SPF_PASS -0.001 SPF: sender matches SPF record Message-ID-Hash: FANGIGNPWUDS43THXS4F3AGF3WZAJAYC X-Message-ID-Hash: FANGIGNPWUDS43THXS4F3AGF3WZAJAYC X-MailFrom: e.huhsovitz@proxmox.com X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; loop; banned-address; emergency; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header X-Mailman-Version: 3.3.10 Precedence: list List-Id: Proxmox VE development discussion List-Help: List-Owner: List-Post: List-Subscribe: List-Unsubscribe: Sensible change! The state information should be allocated as late as possible on disk. Tested using the same methodology as the bug reporter. The state-file was not orphaned and after releasing the QMP socket, operations and snapshots continued working as normal. Therefore: Reviewed-by: Elias Huhsovitz Tested-by: Elias Huhsovitz On Mon Sep 14, 2026 at 4:05 PM CEST, Michael K=C3=B6ppl wrote: > The state volume was allocated before the running instance was queried > for its machine type, CPU argument and network MTUs. Those queries can > fail (e.g. the 'query-machines' QMP command, which can run into a > timeout while QEMU's is busy [0]). > > The helper is called from __snapshot_prepare() before write_config(), so > the config never records the snapshot, snapshot_create() dies before it > gets to its cleanup path. Such a failure leaves behind the just > allocated volume. > > Query the running instance first, which makes the allocation the last > step of the function that can fail. > > [0] https://bugzilla.proxmox.com/show_bug.cgi?id=3D8030 > > Signed-off-by: Michael K=C3=B6ppl > --- > Tested this pretty much as described in the Bugzilla entry, but using a > Perl script to connect to the QMP socket instead of socat. > > src/PVE/QemuConfig.pm | 7 +++++-- > 1 file changed, 5 insertions(+), 2 deletions(-) > > diff --git a/src/PVE/QemuConfig.pm b/src/PVE/QemuConfig.pm > index 26f0fda2..2ef6c075 100644 > --- a/src/PVE/QemuConfig.pm > +++ b/src/PVE/QemuConfig.pm > @@ -238,8 +238,6 @@ sub __snapshot_save_vmstate { > my $name =3D "vm-$vmid-state-$snapname"; > $name .=3D ".raw" if $scfg->{path}; # add filename extension for fil= e base storage > =20 > - my $statefile =3D > - PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $nam= e, $size * 1024); > my $runningmachine =3D PVE::QemuServer::Machine::get_current_qemu_ma= chine($vmid); > =20 > # get current QEMU -cpu argument to ensure consistency of custom CPU= models > @@ -249,6 +247,11 @@ sub __snapshot_save_vmstate { > =20 > my $nets_host_mtu =3D PVE::QemuServer::Network::get_nets_host_mtu($v= mid, $conf); > =20 > + # allocate only after querying the running instance, nothing below c= an fail, so a failed > + # query cannot leave an orphaned state volume behind > + my $statefile =3D > + PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $nam= e, $size * 1024); > + > if (!$suspend) { > $conf =3D $conf->{snapshots}->{$snapname}; > }