+---------[powerDown()] <- Stuck <--[failure]-+
V |
+-> PoweredOff --+-->[powerUp()]--> Starting --+ | +-----[resume()]-----+
| | | | V |
| Aborted <----+ +--> Running --[pause()]--> Paused
| | ^ | ^ |
| AbortedSaved <-----------[failure]-----<----+ | | | |
| | | | | | |
| +-------------+ | | | | |
| V | | | | |
| Saved --------[powerUp()]-----> Restoring -+ | | | |
| ^ | | | |
| | +-----------------------------------------+-|-------------------+ +
| | | | |
| | +- OnlineSnapshotting <--[takeSnapshot()]<--+---------------------+
| | | |
| +-------- Saving <--------[saveState()]<----------+---------------------+
| | |
+-------------- Stopping -------[powerDown()]<----------+---------------------+
Note that states to the right of PoweredOff, Aborted, AbortedSaved, and
Saved in the above diagram are called online VM states. These
states represent the virtual machine which is being executed in a
dedicated process (usually with a GUI window attached to it where you can
see the activity of the virtual machine and interact with it). There are
two special pseudo-states, FirstOnline and LastOnline, that can be used
in relational expressions to detect if the given machine state is online
or not:
if (machine.GetState() >= MachineState_FirstOnline &&
machine.GetState() <= MachineState_LastOnline)
{
...the machine is being executed...
}
When the virtual machine is in one of the online VM states (that is, being
executed), only a few machine settings can be modified. Methods working
with such settings contain an explicit note about that. An attempt to
change any other setting or perform a modifying operation during this time
will result in the @c VBOX_E_INVALID_VM_STATE error.
All online states except Running, Paused and Stuck are transitional: they
represent temporary conditions of the virtual machine that will last as
long as the operation that initiated such a condition.
The Stuck state is a special case. It means that execution of the machine
has reached the "Guru Meditation" condition. This condition indicates an
internal VMM (virtual machine manager) failure which may happen as a
result of either an unhandled low-level virtual hardware exception or one
of the recompiler exceptions (such as the too-many-traps
condition).
Note also that any online VM state may transit to the Aborted state. This
happens if the process that is executing the virtual machine terminates
unexpectedly (for example, crashes). Other than that, the Aborted state is
equivalent to PoweredOff.
There are also a few additional state diagrams that do not deal with
virtual machine execution and therefore are shown separately. The states
shown on these diagrams are called offline VM states (this includes
PoweredOff, Aborted, AbortedSaved and Saved too).
The first diagram shows what happens when a lengthy setup operation is
being executed (such as ).
+----------------------------------(same state as before the call)------+
| |
+-> PoweredOff --+ |
| | |
|-> Aborted -----+-->[lengthy VM configuration call] --> SettingUp -----+
| |
+-> Saved -------+
The next two diagrams demonstrate the process of taking a snapshot of a
powered off virtual machine, restoring the state to that as of a snapshot
or deleting a snapshot, respectively.
+----------------------------------(same state as before the call)------+
| |
+-> PoweredOff --+ |
| +-->[takeSnapshot()] ------------------> Snapshotting -+
+-> Aborted -----+
+-> PoweredOff --+
| |
| Aborted -----+-->[restoreSnapshot() ]-------> RestoringSnapshot -+
| | [deleteSnapshot() ]-------> DeletingSnapshot --+
+-> Saved -------+ |
| |
+---(Saved if restored from an online snapshot, PoweredOff otherwise)---+
"VDI"
"vdi"
"VdI"
refer to the same medium format.
Note that the virtual medium framework is backend-based, therefore
the list of supported formats depends on what backends are currently
installed.
--vminfo-user-idle-threshold <ms>, or by setting the per-VM guest property
/VirtualBox/GuestAdd/VBoxService/--vminfo-user-idle-threshold <ms>with the RDONLYGUEST flag on the host. In both cases VBoxService needs to be restarted in order to get the changes applied.
BEFORE attaching B.vdi: AFTER attaching B.vdi:
Snapshot 1 (B.vdi) Snapshot 1 (B.vdi)
Snapshot 2 (D1->B.vdi) Snapshot 2 (D1->B.vdi)
Snapshot 3 (D2->D1.vdi) Snapshot 3 (D2->D1.vdi)
Snapshot 4 (none) Snapshot 4 (none)
CurState (none) CurState (D3->D2.vdi)
NOT
...
CurState (D3->B.vdi)
The first column is the virtual machine configuration before the base hard
disk B.vdi is attached, the second column shows the machine after
this hard disk is attached. Constructs like D1->B.vdi and similar
mean that the hard disk that is actually attached to the machine is a
differencing hard disk, D1.vdi, which is linked to (based on)
another hard disk, B.vdi.
As we can see from the example, the hard disk B.vdi was detached
from the machine before taking Snapshot 4. Later, after Snapshot 4 was
taken, the user decides to attach B.vdi again. B.vdi has
dependent child hard disks (D1.vdi, D2.vdi), therefore
it cannot be attached directly and needs an indirect attachment (i.e.
implicit creation of a new differencing hard disk). Due to the smart
attachment procedure, the new differencing hard disk
(D3.vdi) will be based on D2.vdi, not on
B.vdi itself, since D2.vdi is the most recent view of
B.vdi existing for this snapshot branch of the given virtual
machine.
Note that if there is more than one descendant hard disk of the given base
hard disk found in a snapshot, and there is an exact device, channel and
bus match, then this exact match will be used. Otherwise, the youngest
descendant will be picked up.
There is one more important aspect of the smart attachment procedure which
is not related to snapshots at all. Before walking through the snapshots
as described above, the backup copy of the current list of hard disk
attachment is searched for descendants. This backup copy is created when
the hard disk configuration is changed for the first time after the last
call and used by
to undo the recent hard disk
changes. When such a descendant is found in this backup copy, it will be
simply re-attached back, without creating a new differencing hard disk for
it. This optimization is necessary to make it possible to re-attach the
base or immutable hard disk to a different bus, channel or device slot
without losing the contents of the differencing hard disk actually
attached to the machine in place of it.
<path>/{<uuid>}.<ext>
where <path> is the supplied path specification,
<uuid> is the newly generated UUID and <ext>
is the default extension for the storage format of this hard disk. After
that, you may call any of the methods that create a new hard disk storage
unit and they will use the generated UUID and file name.
Base <- Diff_1 <- Diff_2Here, calling this method on the Base medium object with Diff_2 as an argument will be a forward merge; calling it on Diff_2 with Base as an argument will be a backward merge. Note that in both cases the contents of the resulting medium will be the same, the only difference is the medium object that takes the result of the merge operation. In case of the forward merge in the above example, the result will be written to Diff_2; in case of the backward merge, the result will be written to Base. In other words, the result of the operation is always stored in the target medium. Upon successful operation completion, the storage units of all media in the chain between this (source) medium and the target medium, including the source medium itself, will be automatically deleted and the relevant medium objects (including this medium) will become uninitialized. This means that any attempt to call any of their methods or attributes will fail with the "Object not ready" (E_ACCESSDENIED) error. Applied to the above example, the forward merge of Base to Diff_2 will delete and uninitialize both Base and Diff_1 media. Note that Diff_2 in this case will become a base medium itself since it will no longer be based on any other medium. Considering the above, all of the following conditions must be met in order for the merge operation to succeed:
"VDI"
"vdi"
"VdI"
refer to the same medium format.
This string is used in methods of other interfaces where it is necessary
to specify a medium format, such as
.