A shut-off clone kept the VM's TPM element as it was: an emulated TPM
with a state path of its own, a host's TPM or an external one would be
shared with the VM, not the new, empty TPM the clone form promises. An
emulated TPM loses its state path, and the others are refused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Kk4h2TcUiaWmU5MUszYSea
libvirt refused the copy only after the job had begun; recovery then
found a volume of the planned name and ended the job uncertain, holding
the VM. The names are checked when the job is queued and before it runs,
as a live clone did; that check now refreshes the pool first, so a live
clone's free space is read current too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Kk4h2TcUiaWmU5MUszYSea
Only a file in libvirt's NVRAM directory was copied; one elsewhere was
skipped without a word, and the VM started on the destination from the
firmware's template, its saved boot entries lost. Such a VM is now
refused before anything changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Kk4h2TcUiaWmU5MUszYSea
A worker that stopped during the copy left the partial files on the
destination: libvirt ends the migration and the VM stays, but the files
took the space and refused the next try, as a copy never replaces a
file. The job records the copies before libvirt makes them, and
recovery deletes them while the destination has no VM of it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Kk4h2TcUiaWmU5MUszYSea
A copied disk was found by its target as well as its path: after a bus
change waiting for the next start (vda to sda), the next start's
definition kept the source path, which the destination does not have,
and the VM would not start again there. The path alone finds it, so two
disks of one file can no longer be copied to two places.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Kk4h2TcUiaWmU5MUszYSea
The source host alone was asked: a file on storage the computes share,
used by a VM or volume of another compute, was deleted under it. The
check the disk move makes, every compute and a host that cannot tell
counting as using it, moves to instances.utils and serves both. The
destination is not asked when its copy has the source's path: its VM
uses that copy, another file.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Kk4h2TcUiaWmU5MUszYSea
Each host's logical volumes are its own: a VM whose disk is /dev/vg/x
migrated, uncopied, onto the destination's volume of the same name and
kept running on other data. Such a volume now counts as missing there.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Kk4h2TcUiaWmU5MUszYSea
A disk of the user's VMs whose size could not be read (its volume not in a
pool, the pool stopped) counted as 0 GiB, so a disk-limited user's increase
could pass over the limit. A disk no pool holds now has the size libvirt
reads from the file or device (blockInfo, also for a shut-off VM); one whose
size is still unknown makes the usage unverified and the increase is
refused, as when a host cannot be reached. Users without a disk limit are
not affected.
A running VM's page read its addresses while it rendered: the guest agent,
when its channel is connected but it does not answer, holds the page (and
a gunicorn worker) for its whole 10 s timeout, on every load. The page now
shows them once it has loaded: vm_addresses.js asks the new
instances/<pk>/addresses/ for each NIC's addresses and fills the header's
IPv4 and each network card's IPv4 and IPv6.
- A migration whose libvirt call failed but which recovery finds moved (or
cannot tell) gets a row of that end; only a failure already logged by
migrate_instance is not logged twice.
- A clone made with its resize or start failed, and a recovered migration
whose autostart or NVRAM step failed, write a failed row of that step, as
migrate_instance does for its later steps.
- Setting the link state of an interface or the video model of a VM that
has none is refused: it changed nothing and was logged as done.
- The rows of a removed VM record (deleted or expired) name its host, VM
and UUID, so the VM's history has them.
Since the VM create became a job (6f0c7300) every new VM was defined with
no VNC password, and a clone copied its source's password (or its lack
of one) exactly: every clone of a template shared it. Computes listen
for VNC on all their addresses (dev/libvirt-bootstrap.sh), so anyone who
could reach a compute's VNC ports opened any of these consoles directly,
without the panel; the README's firewall rule was the only guard.
The create job asks for create.py's random 8-character password again,
and a clone's VNC graphics get a new random one. The console page sends
the password to a user with passwordless_console, which the user create
form now ticks by default next to change_password; a user without it is
asked for the password, as the permission intends. Existing VMs keep
their settings.
Live create, clone and VM page tests pass on a TCP compute.
The delete ran in the request: undefine, then the VM's record, then its
disk files. A request that died in between (a gunicorn worker past its
timeout, OOM, a restart), or a database write that failed after the
undefine, left a record that later showed as removed outside the panel,
and the disk files the user asked to delete were never deleted.
Deleting a VM is now an instance.delete job. It holds the compute
exclusively, as the request did, so no other VM attaches a file it is
about to delete. The job chooses the files before it stops the VM (a host
that cannot be checked still stops nothing) and records them first. It
then stops and undefines the VM, removes its record, and deletes the
files, each with its own Logs row. Recovery decides from libvirt: a VM
still on its host means the delete failed; a VM gone means libvirt did
it, so recovery removes the record and deletes the recorded files that
are still there.
The page queues the job and returns to the VM page, where the job panel
follows it and opens the VM list when it ends; disks that were kept are
the result's warning. The API's delete answers 202 with the job and
needs a submission_key. backing_path moves to wvmConnect, as recovery
has no domain to ask.
Each poll called info(), then blockStats() for every disk and
interfaceStats() for every NIC: nine calls for a VM with four disks and
two NICs, every 5 s for each open Stats tab. domainListGetStats now gives
the CPU time and every disk's and NIC's counters at once; a counter the
host does not report is -1, as before.
The internal revert's shut-off check ran in the job's run, after its
intent was recorded, so the refusal went to recovery: the job ended
uncertain and kept the VM's claim until a superuser resolved it. The check
(check_internal_revert) now runs in precheck too, so a page older than the
VM's start, or a direct POST, fails the job with the reason.
libvirt requires a stream that did not finish to be aborted before it is
freed. The NVRAM download of a shut-off move and upload_volume (NVRAM
upload, cloud-init seed) freed theirs after a failed send, receive or
finish without aborting, on the shared host connection. They abort now;
upload_volume opens its stream before it makes the volume, so the
cleanup has both.
The Resize tab offered vCPU counts only up to the host's logical CPUs.
A VM whose maximum was higher (4 on a 2-CPU host, say) found no option
for it, so the Maximum list showed 1, and a resize of the shut-off VM
lowered its maximum to that without a word.
get_max_cpus() now takes the VM's own maximum as a floor, and the
Instance's vcpu_range passes it. The host's CPU count shown on the page
and used by the clone form is read from the host alone, not from that
range.
wvmInstance kept the guest agent's and the ARP table's addresses from
the first NIC it was asked about, and never read them again on that
object; it carried a lease lookup that was never filled.
The addresses are now one {MAC: (IPv4, IPv6)} read, kept like the
domain XML only inside a cached_reads block: the VM page, which reads
its NICs twice, asks the agent once (it may take its whole timeout),
and a read outside a page's block is a new one. The agent's addresses
still come before the ARP table's.
vrtManager's errors (some 140: a disk that does not move, a network
address outside its subnet, a host device given to another VM, the
disks a migration cannot copy) were English f-strings, so no catalog
could translate them, on pages or in jobs.
They are marked texts with their values now; a variant that differs by
a word (internal or external, running or persistent, with a subnet or
without) is a whole message of its own, and a list of problems is a
list of messages. OperationError reads in the active language, so the
pages that show one translate it as they are; its args stay English, and
so do the logs and the audit log. Host device problems and the Migrate
tab's reasons a disk is not copied are messages too.
Two disks of a VM with the same file name in different directories
were both offered the same destination pool by default. The form was
sent, and the migration job refused it only then: both copies would
be one file.
The plan now gives each such disk, by default, a pool no earlier disk
of that name is copied into, and says under its choice that another
disk has the file name. A choice that still puts them in one pool is
refused by the job, as before.
A pool whose refresh failed, one busy with a volume copy for example,
was skipped without a word, so a hand-copied image stayed missing with
no reason given. The other pools are still refreshed; each failure is
shown with libvirt's error.
The API comment no longer says listing the ISO images refreshes pools.
The VM page refreshed every active storage pool to list the ISO images
for its CD-ROM drives. A pool holding large disk images took up to
seconds to refresh, most of the page's time, on every page load.
The page now lists the ISO images libvirt already knows. Images
uploaded through the panel are known at once. For an image copied to a
pool by hand, the Removable media section has a Refresh ISO list button
that refreshes the active pools and returns to the page.
The Migrate tab shows a running VM's live migration switch, on by
default. A VM shut down after the tab opened, as one is to move it shut
off, was then refused: "A shut-off VM cannot be migrated live". A
shut-off VM now moves offline whatever the request says about live, as
the tab tells for a shut-off VM; the tuning options are for a running VM
only, as before. An offline request for a running or paused VM is still
refused.
A shut-off VM with TPM state, as every Windows 11 VM, was refused for
migration: the state is a file of the source host, and an offline move
would start it there with a new TPM. Such a VM now migrates as a shut-off
VM with disks to copy does: its UEFI variables are copied, it is started
paused, its CPUs never run, migrates live, which carries the TPM state,
and is shut off on the destination. A shut-off VM with a saved state is
refused before any of that, for every way it moves: started paused, it
would resume and then lose the saved memory.
The clone tab says that a clone gets a new, empty TPM, and the destroy
page that the TPM state goes with the VM.
A VM made without a TPM, a Windows 10 to be upgraded to Windows 11, had
no way to get one but the XML editor. The Options tab has a TPM 2.0
switch, as the RNG one: it adds an emulated TPM of the host's model to
the definition for the next start, or removes it. Its state stays on the
host until the VM is undefined, so a TPM added again has it back. A TPM
the switch did not make (passthrough, 1.2, encrypted, with its own state
path) is left to the XML editor, and a host without swtpm says so.
A running VM got its new CD-ROM on SCSI, the one bus the panel could plug
in then, and Windows does not see a virtio-scsi drive before the driver
on that ISO is installed. Add CD-ROM is now a dialog that chooses the bus
and the ISO: a running VM takes USB, first, which every guest sees and
which hot-plugs, or SCSI; a shut-off one SATA (IDE on pc), SCSI or USB.
libvirt has no empty USB drive, so a USB CD-ROM takes its ISO when it is
added and is detached instead of ejected, also while the VM runs.
A USB CD-ROM is a CD drive on QEMU's usb-bot only. Where QEMU has none
(RHEL 9), libvirt took usb-storage, which shows the guest a hard disk
that Windows does not read; the drive now asks for usb-bot, and such a
host refuses it with libvirt's message.
An existing q35 VM whose PCIe root ports are all used takes no new
virtio disk or network device while it runs, and the only way to give
it more was editing its XML.
The Disks tab of a PCIe machine now counts its slots, "PCIe slots: 0
free of 7", in red when none is free, with what its next start will
have when that differs, and Add PCIe slots adds four root ports to its
persistent definition: libvirt numbers and places them, and a running
VM gets them at its next start. The no-slot advice of Add disk and Add
network device points at it.
The Edit dialog of a VM's network device selected the listed network
or interface whose name matched the device's source, and the first one
when none did: saving only a new model moved a macvtap device or a
bridge the host's lists did not show onto the first network, a bridge
named like a network became that network, and a macvtap device chosen
again lost its mode.
The dialog now offers the device's own source first, "<name>, as it
is", and posts nothing for it; the view passes no source and the
device's source element is left as it is. Choosing another source
changes it as before.
The Migrate tab offered every other host alike, and its plan showed
neither a file of the same name in a destination pool nor the disks of
the VM's next start, which the job then refused. Now:
- When the tab opens, the hosts are checked first: one that cannot be
reached, or the VM's own host under another record (the same libvirt
host UUID), is shown with why and cannot be chosen, and the disks of
the first one left are checked. The job refuses the VM's own host
under another name too.
- A pool of the destination that has a file of a disk's name is shown
but cannot be chosen for it; a disk no pool can take refuses the form.
- The next start's file disks the migration does not carry are listed;
one not on storage both hosts share refuses the form, as the job does.
Destroying a VM with its disk files forced it off first and then
worked out which files to delete, so a check that failed left the VM
stopped, and only this host's VMs and volumes were checked: a file on
shared storage that a VM on another host uses was deleted.
The files are now worked out before the VM is stopped: the disks it
runs with and those of its next start, as before with their snapshot
files and seed. A file a VM or volume of another of the panel's hosts
uses at the same path (shared storage is mounted at one path on every
host, as migration needs) is kept like one used here; another record
of the VM's own host is skipped, as there the VM itself uses them. A
host that cannot be read stops the destruction with its name, before
anything is stopped or deleted.
A user's quota counted each VM's running vCPUs, memory and disks, so a
larger value set for the VM's next start was not counted, and a clone
counted only once its VM was recorded, so a user could queue clones
past the quota. Now:
- Each VM counts the larger of its running and its next-start values.
- A clone job counts as one VM with the source's persistent vCPUs,
memory and disk size, which its plan stores, from its admission until
its VM is recorded; a failed, cancelled or resolved job no longer
counts. The clone's VM and its owner are recorded in one transaction,
so the quota counts either the job or the VM, never neither.
- The quota reads each VM's definition once (cached_reads).
The Stats tab stacked one large card per chart, so a VM with a few
devices made a long page, and its "Real Time" and "Logs" sub-tabs were
shown to every user although only a superuser may read the logs. The
owner chose the layout from rendered options: a state line, CPU and
memory side by side, one chart per NIC and disk below, each with its
latest reading next to its title, and a small Graphs/Logs switch for
superusers.
Codex reviewed the tab; this is fixed:
- Memory is shown in MiB (it was labelled MB), and its axis grows to
the RSS when the RSS passes the balloon size instead of clipping it.
- The CPU reading says it is of the host's CPUs.
- A NIC is titled by its MAC, not an invented "eth0", and its readings
are matched to its chart by the MAC: they were matched by position, so
removing a NIC while the page was open moved the next one's readings
into its chart.
- A VM that is not running says nothing is measured.
- The logs load whenever the Logs view is shown, by keyboard too; an
empty answer says "None" and a failed one says so, instead of an
empty table. The tab shows the latest 50 rows, with a link to all of
the VM's rows on the Logs page.
- The VM's header shows memory in MiB; it said MB.
The Destroy tab was a sub-tab with one sentence and a link; its
confirmation page said only "This action is irreversible!" over a
"Remove Instance's data" switch. The tab now shows the state and
"Review destruction". The owner chose the confirmation layout from
three rendered options: the VM and its host, that its definition,
saved state and snapshot records go and that a running VM is forced
off, two plans for the files (delete the disk files, or keep them), the
VM's disks, what is always kept, the NVRAM choice, and Cancel.
Codex reviewed destroying; this is fixed:
- A disk file that cannot be deleted keeps the files below it, which
it needs: they were deleted after it, leaving it broken. The files
are deleted overlays first (split_disk_paths_by_use orders them by
how many files each grows from), and if the files a kept one needs
cannot be read, none of the rest goes. A message names each file
kept so, and a failed deletion says the VM is removed.
The Snapshots tab had three sub-tabs: internal, external and the
lists, with the two kinds side by side as equals. The owner set the
direction: external snapshots are the way; internal ones pause the VM
while its memory is saved and can take long, so they stay only as a
fallback for a VM that cannot take an external one.
The tab now has the external snapshot form on top, its name and
description on one line, with its memory and freeze options; the
internal snapshot under a folded "fallback" line with its warning; and
one list of all snapshots, newest first, each on one line with its
name, description, kind and date, and Revert and Delete.
Codex reviewed snapshots end to end; these are fixed:
- The pause and memory lines follow the VM's state and the kind.
- An internal snapshot is reverted to on a shut-off VM only, on the
server too; the tab alone held that rule.
- A paused VM's revert says it comes back paused, not that it starts.
- An external snapshot waits, with the reason, while snapshots newer
than the one reverted to exist; the server refused it unexplained.
- Deleting an external snapshot says its files are merged with the
disks; "into the next one" was wrong for the last one.
- The description has no 45-character limit only the browser kept.
- Freezing is not shown ticked where it cannot be offered, and
unticking memory gives back the earlier freeze choice.
- Internal create, revert and delete refuse a saved state or a running
disk job, as external ones do.
The Resize tab's CPU, Memory and Disk panes were horizontal forms with
bold host figures. The owner chose the ledger layout from three
rendered options: each section with its heading, the VM's figures and
the host's on the left, its fields on the right, and one line on what
applies now and what at the next start.
Codex reviewed resizing end to end; these are fixed:
- A vCPU turned on or off while the VM runs is kept at its next start:
libvirt was asked without flags, which changes the running VM only.
- A paused VM offers to turn vCPUs on only: turning one off needs the
guest, which does not run.
- A smaller disk size is refused, and so is the rest of the form,
before anything grows; it was skipped without a word.
- vCPUs and memory below 1 are refused.
- The memory fields are numbers filled with the VM's values: the
"Custom value" box, its script and the preset list are gone; an empty
custom field silently used the preset.
- Units and words: MiB, not MB; a disk's field is its new size, with
its size now above it; the messages say what the VM gets.
Two settings the owner chose to add to Settings > Options after the
page's review with Codex. Both change the persistent definition, so
they are saved while the VM runs and used from its next start.
- Random number generator: a switch for a virtio RNG device fed from
the host's /dev/urandom, so a new VM does not wait for entropy at
boot. Another RNG device, or the panel's with a rate or driver, is
shown as the VM's own and left to the XML editor.
- When the guest crashes: Restart or Stop (on_crash restart or
destroy); libvirt's default, Stop, when the definition has none.
Stop keeps a VM that crashes again and again down for a look. Any
other action is shown and left to the XML editor.
Settings > Options was four forms under bold sentences, most of them
disabled while the VM ran. The owner chose this layout from three
rendered options: one list, each row a setting with its description on
the left and its control on the right: VM details (title, description),
Template, primary video model, QEMU guest agent. vCPU hot plug moved to
Resize > CPU, next to the vCPU counts.
Codex reviewed the four forms; these are fixed:
- Title, description and the video model are saved while the VM runs:
they are in the persistent definition, used at the next start.
- The template flag has its own form and view, set_template, for
superusers and staff, and only for a shut-off VM: a running VM made a
template left the quota while it kept running. Turning a template
back into a VM checks the quota there, and each change is audited.
The title form no longer touches the flag.
- A new video model stays the primary one; the old model's memory and
head settings, which are its own, go with it.
- The agent and vCPU hot plug forms show the persistent definition; they
showed the running VM's, so a pending change looked undone.
- Enabling the guest agent adds its channel only where it is missing:
saving Enabled again failed, and a channel pending the next start kept
a running VM from getting one.
- The agent line tells a VM that is not running, one without a channel,
and a connected or not connected agent apart.
Settings > Clone was a horizontal form with unlabelled checkboxes. The
owner chose this layout from three rendered options: the clone's name
on top, a "Network" card with a new MAC address per NIC, a "Storage"
card with a new volume name per disk, then the optional title and
description and a footer with Clone.
Codex reviewed cloning end to end; these are fixed:
- Every NIC of the clone gets a new random MAC when the form opens.
Only the first did, so a VM with two NICs gave its clone the second
NIC's MAC. Guess MAC is offered for the first NIC only: a name's DHCP
host entry holds one address.
- A VM without a NIC can be cloned: the first NIC's MAC was required.
- The disk name script uses the persistent disks, as the form does; a
disk added pending the next start kept the source's volume name.
- A running or paused VM is not cloned while a copied disk changes at
the next start: the backup copies the running disk, so the clone
would get the old contents under the new definition.
- A derived volume name that does not start with the source's name now
starts with the clone's name and the disk's device, so a second clone
or a second disk does not take the same name.
- Guess for the clone name says when the DHCP file has no free name; it
stopped with a script error.
- The qcow2 metadata preallocation box shows only for a shut-off VM: a
live clone's backup ignores it.
Other users no longer get hidden MAC and disk inputs or the scripts that
filled them: the server derives those values and ignored the posted ones.
Settings > Console was one horizontal form, and on a running VM its
Update button was disabled. The owner chose this layout from three
rendered options: the console type on top, then a "VNC connection" card
(listen address, keymap) and a "VNC password" card.
- The password is changed by an action: keep it, generate a new one,
set a new one or remove it. This replaces the Generate and Clear
boxes and their precedence, and an emptied field no longer removes
the password silently. The pane says whether a password is set; the
form no longer carries the password itself.
- A VNC password takes at most 8 bytes: libvirt refuses a longer one
("VNC password is 9 characters long, only 8 permitted"). Generate and
a new VM's random password now make 8 characters; they made 12, which
libvirt 11 refuses. A set password keeps its spaces.
- Every check is in the form, so a refused save changes nothing; before,
a refused password removal still saved the other settings.
- The settings can be saved while the VM runs: they are in the
persistent definition. A line says the running console keeps its
settings until the VM is shut down and started.
- The console type shown and compared is the persistent one; it was the
running VM's, so a pending switch to Serial showed VNC and a later
save could switch again.
- While the saved type is Serial, the VNC settings are not shown or
taken; choosing Serial ignores the posted VNC settings.
The Boot pane shows Autostart and the boot menu as switches that save on
change, and the boot order as one list of the VM's disks, CD-ROM drives
and NICs: the devices in the order first, checked and numbered, the
others below. Up and Down move the checked devices; Apply order saves.
- The boot menu can be changed while the VM runs: it is in the
persistent definition, so it applies at the next start, as the boot
order does.
- The boot order editor no longer needs the boot menu enabled.
- A VM that boots by device type (<os><boot dev='hd'/>) names those
types; checking devices replaces that order.
- An order is refused while a device the list does not show (a host
device, a floppy) has a boot place, so Apply does not drop it; the
message points to the XML editor.
- An empty order is refused with a message instead of being ignored.
- get_bootorder gives a NIC's full MAC as its target.
- The unused jQuery multiselect plugin is removed.
Settings > Network was a table whose fields were read-only inputs cut
off at phone width, an unlabelled "+", three icon buttons per NIC and
a separate QoS table keyed by MAC. The owner chose card layout from
three rendered options, with the QoS limits folding out like the
device details:
- A card per NIC: a VM's only NIC spans the row, two or more sit side
by side from md. Each has its name and target, the link state as a
switch, its source and IPv4 addresses, and "Device details"
(type, MAC, model, NWFilter, IPv6) in a <details>.
- "QoS limits" in a second <details>, its summary naming the limited
directions: a form per inbound or outbound limit, with Save and
Delete limit, and the units.
- Labelled Edit, Add QoS and Delete in the card footer; "Add network
device" by the heading. The edit and add-QoS dialogs are titled and
labelled, and the edit dialog's fields have labels.
The limits come from each NIC of the persistent definition, so the QoS
table, Instance.config_qos and wvmInstance.get_all_qos are removed.
The Disks tab has a Move button for each disk: it opens a dialog to
choose another storage pool of the same host, and the move runs as a job.
- Running or paused VM: libvirt copies the disk with a block job
(TRANSIENT_JOB, as the VM is persistent) into a file it creates in the
pool, then pivots. The pivot names the copy in both the running and
the next start's definition.
- Shut-off VM: the volume is copied into the pool and the copy named in
the definition.
The source file goes then, unless a VM or volume of any compute uses it
(shared storage holds it for every host that mounts it), a compute
cannot be read, or the delete fails: the job lists such a file under
"Left behind", and the move stands.
A disk moves when:
- the VM is defined and running, paused or shut off, without snapshots,
a saved state or a block job;
- the disk is a writable, unshared, unencrypted raw or qcow2 file of a
storage pool, without a backing file, and the same file in both
definitions;
- the pool keeps disk files, is another directory, and has no file of
that name (libvirt would write into it).
Otherwise it is refused with the reason before anything changes.
The job claims the VM and the file's name in both pools. A cancel aborts
the block copy, and ends the job cancelled once the VM is back on its
file. A failed move deletes the copy it made, unless a definition names
it (a lost reply after the pivot); a file there before the copy began is
never deleted. Recovery after a stopped worker
ends a copy left mirroring, matched by its destination, and reads the
outcome from the definitions.
Superusers only, like the other disk edits.
A migration needed every disk on storage both hosts share: it was refused
when the destination lacked a disk file, and a file of the same name
there was taken for the VM's own disk.
The Migrate tab now lists the VM's disk files for the chosen destination.
For each disk it offers the destination's pools that keep disk files
(directory, filesystem and NFS pools). A disk on storage both hosts share
can also stay where it is.
A directory counts as shared when both hosts have it as a storage pool
and a marker file created on the source exists on the destination. A
directory the probe cannot tell is never taken as shared.
libvirt copies the chosen disks during the migration: migrate3 with
NON_SHARED_DISK for those targets, and destination and persistent
definitions naming the new paths, where libvirt creates the files.
- A shut-off VM is started paused for the copy and shut off on the
destination; its CPUs never run. Its UEFI variables are copied first,
as the destination would start from the firmware's template. If the
migration fails, the VM is shut off on the source again.
- Refused before anything changes, each disk with its reason:
- a disk neither shared nor given a pool, also one only the next
start's definition has;
- CD-ROM images, read-only or encrypted disks, formats other than raw
and qcow2, and disks with a backing file;
- a pool that is not of disk files, or that has a file of that name,
which libvirt would reuse and overwrite;
- two disks copied to one file.
- The job claims each copy's name in its pool, and the source volume.
- A failed or cancelled migration deletes the copies libvirt made, and
the UEFI variables copy, unless a VM on the destination has them.
- The source files stay and are named in the log. With "Delete the
source files of copied disks" they go, except one another VM or volume
uses; a failed delete is logged and the migration stands.
The NVRAM sharing check uses the same probe: a shut-off move stops when
it cannot be told, and the source file stays.
The API's migrate call takes disk_pools ({target: pool}) and
delete_source.
- A running VM gets a PCI device of its host at once and for its next
start, in one libvirt call; a hot plug libvirt could not keep for the
next start says that the device is in the running VM only. A q35 VM
without a free pcie-root-port gets libvirt's error and where a slot
comes from.
- A PCI device is taken from the running VM and from its next start, each
as it has it; a refusal by the guest stops there. When the guest has
not let the device go within libvirt's wait, the page says the removal
is requested and the Logs row records the request: the device stays in
the running VM until the guest lets it go or the VM stops.
- A paused VM, or one with a saved state, does not change its PCI devices.
- A device for a running VM is checked alone too, as libvirt takes only
the hot-plugged device from its host driver.
- The Host devices tab gives a shut-off VM a PCI device of its host and
takes it back; libvirt binds the device to vfio-pci when the VM starts
and gives it back when it stops (managed). The VM's PCI devices are
listed with whether the running VM and its next start have them.
- A device is offered, and accepted on the server under the compute's
claim, only when it is no bridge of any kind (bridge header, PCIe port,
or of the bridge class: host bridge, LPC), the host has none of its own
network interfaces, disks or graphics card on it (else: bind it to
vfio-pci on the host first), no other VM names it, and every other
endpoint of its IOMMU group is the VM's too or has no driver or
vfio-pci and no other VM; bridges in the group stay with pcieport, and
a member that cannot be read keeps the group. Nothing in the group is
taken implicitly.
- A snapshot revert checks the PCI devices of the snapshot with the same
rules; migration and clones already leave host devices out.
- The add forms of the tab load together, when the tab opens.
- A Host devices tab on the VM page lists the VM's USB host devices, each
once: how it is chosen (its address, or any device of its model), the
device the running VM holds and whether its next start has it; its other
host devices (PCI, SCSI, mediated) are listed for the XML editor.
- Superusers add a USB device of the host and take it back, on a running
VM too. The add form is loaded when the tab opens, as its choices read
every VM of the host; a hub, a device the host has its own disks or
network interfaces on, or one another VM names is not offered, with the
reason, and refused on the server under the compute's claim. A device is
chosen by model when the host has one of it, else by its address.
- A VM without a USB controller is refused; a device is taken from each
definition as that definition describes it.
- A clone gets none of the VM's host devices (the clone form says so), and
a VM with host devices is not migrated. A snapshot revert is refused when
it would give the VM a device another VM names or the host uses.
- A NIC that is a host device is no longer edited or removed from the
Network tab: the XML editor changes it.
Plan reviewed with Codex; libvirt's behaviour checked on a host.
- A filter is not deleted while a VM's interface references it, as the
VM runs or as saved for its next boot (a stopped VM too), or while
another filter references it: libvirt refuses only one a running VM
uses, and a VM using a filter that references a deleted one does not
start. The VMs are read from one listing, one or two XML reads each,
instead of a connection per VM; the refusal names them.
- Create refuses a name or UUID the host has, clone a name in use or
one outside the panel's names, and editing the XML must keep the
filter's name and UUID: each would replace another filter or make a
copy. Entered XML is read for these checks without expanding entities
or taking a DTD (util.parse_xml, shared with the network editor),
and a rule is one <rule> element.
- A rule or reference is deleted only while it reads as the page showed
it, not by index and three attributes: a rule alike in those or a
filter changed meanwhile loses nothing.
- A reference is added only to another filter of the host not
referenced yet; the picker offers those.
- References keep their parameters, in order (a name may repeat); a
rule shows every protocol element it has, a rule without one too.
- One change a POST (two would both start from the page's reading,
the second undoing the first); every change redirects and reports its
refusal as a message, a POST for a filter gone meanwhile going to the
list; the catch-all that showed any error goes, and so does
wvmInstance.get_filterrefs, which only the old delete check used. The list reads each filter's
XML once from one listing; a filter's page reads its XML once and
only the other filters' names; the per-filter XML dialog of the list
goes (a filter's page shows its XML). The XML editors send their text
untouched too.
A VM defined with QEMU's own VNC websocket (graphics websocket='...',
which the panel never sets) made the console page connect to that port on
the panel's host name: the port is on the compute, so the console did not
open unless the panel ran on that compute, and then it bypassed novncd and
the panel's console token. Such a VM has its ordinary VNC port too, and
novncd now reaches it there like any other VM.
When the guest agent failed, the Guest Info section stayed empty with no
word: the request's error answer was not handled. An agent without the
time zone or host name command lost the OS details it had answered too.
The section now shows the agent's error, and a missing command leaves out
only its own details.
Only external snapshots had their name checked. An internal one could be
taken with an empty name: libvirt named it, the job and its log row named
nothing, and a recovery looked for "" and marked the job failed though the
snapshot existed. Both kinds now pass the same check when queued, and the
internal form asks for a name as the external one does.
Snapshot dates showed no year, so last year's snapshot looked like this
year's in the list one reverts from; they now read 2026-10-13 23:04:00
under "Date (time zone)", as on the Jobs page.
Cloud images read their first-boot settings with cloud-init. The panel
served them only over HTTP, which a VM reaches only with the network
and a reverse DNS name for its address. A new VM can now carry them
itself: Cloud-init seed (Advanced) on the create form, or cloud_init in
the API.
The job writes a NoCloud seed, an ISO labelled cidata, into the disk's
directory pool as cidata-<uuid>.iso and puts it in the VM's CD-ROM
drive. Its meta-data holds the VM's UUID as instance-id and its name as
hostname, written as JSON so a VM named 123 or yes keeps a string
hostname; its user-data is what the HTTP datasource serves, with the
creating user's SSH keys. The ISO is built with pycdlib and written
through a libvirt stream, as the NVRAM copy is, now one helper for both.
A seed needs a CD-ROM bus and a directory pool, else the create is
refused. A failed create deletes the seed with the new disk, recovery
names one left behind, and deleting the VM with its disks deletes its
seed, found in the pools by name even when ejected. A clone's drive is
left empty: the seed is the source VM's identity.
