core-admin/doc/qubes-storage.rst

155 lines
8.2 KiB
ReStructuredText
Raw Permalink Normal View History

:py:mod:`qubes.storage` -- Qubes data storage
=============================================
Qubes provide extensible API for domains data storage. Each domain have
multiple storage volumes, for different purposes. Each volume is provided by
some storage pool. Qubes support different storage pool drivers, and it's
possible to register additional 3rd-party drivers.
Domain's storage volumes:
- `root` - this is where operating system is installed. The volume is
available read-write to all domain classes. It could be made read-only for
:py:class:`~qubes.vm.appvm.AppVM` and :py:class:`~qubes.vm.dispvm.DispVM` to
implement an untrusted storage domain in the future, but doing so will cause
such VMs to set up a device-mapper based copy-on-write layer that redirects
writes to the `volatile` volume. Whose storage driver may already do CoW,
leading to an inefficient CoW-on-CoW setup. For this reason, `root` is
currently read-write in all cases.
- `private` - this is where domain's data live. The volume is available
read-write to all domain classes (including :py:class:`~qubes.vm.dispvm.DispVM`,
but data written there is discarded on domain shutdown).
- `volatile` - this is used for any data that do not to persist. This include
swap, copy-on-write layer for a future read-only `root` volume etc.
- `kernel` - domain boot files - operating system kernel, initial ramdisk,
kernel modules etc. This volume is provided read-only and should be provided by
a storage pool respecting :py:attr:`qubes.vm.qubesvm.QubesVM.kernel` property.
Storage pool concept
--------------------
Storage pool is responsible for managing its volumes. Qubes have defined
storage pool driver API, allowing to put domains storage in various places. By
file-reflink, a storage driver optimized for CoW filesystems This adds the file-reflink storage driver. It is never selected automatically for pool creation, especially not the creation of 'varlibqubes' (though it can be used if set up manually). The code is quite small: reflink.py lvm.py file.py + block-snapshot sloccount 334 lines 447 (134%) 570 (171%) Background: btrfs and XFS (but not yet ZFS) support instant copies of individual files through the 'FICLONE' ioctl behind 'cp --reflink'. Which file-reflink uses to snapshot VM image files without an extra device-mapper layer. All the snapshots are essentially freestanding; there's no functional origin vs. snapshot distinction. In contrast to 'file'-on-btrfs, file-reflink inherently avoids CoW-on-CoW. Which is a bigger issue now on R4.0, where even AppVMs' private volumes are CoW. (And turning off the lower, filesystem-level CoW for 'file'-on-btrfs images would turn off data checksums too, i.e. protection against bit rot.) Also in contrast to 'file', all storage features are supported, including - any number of revisions_to_keep - volume.revert() - volume.is_outdated - online fstrim/discard Example tree of a file-reflink pool - *-dirty.img are connected to Xen: - /var/lib/testpool/appvms/foo/volatile-dirty.img - /var/lib/testpool/appvms/foo/root-dirty.img - /var/lib/testpool/appvms/foo/root.img - /var/lib/testpool/appvms/foo/private-dirty.img - /var/lib/testpool/appvms/foo/private.img - /var/lib/testpool/appvms/foo/private.img@2018-01-02T03:04:05Z - /var/lib/testpool/appvms/foo/private.img@2018-01-02T04:05:06Z - /var/lib/testpool/appvms/foo/private.img@2018-01-02T05:06:07Z - /var/lib/testpool/appvms/bar/... - /var/lib/testpool/appvms/... - /var/lib/testpool/template-vms/fedora-26/... - /var/lib/testpool/template-vms/... It looks similar to a 'file' pool tree, and in fact file-reflink is drop-in compatible: $ qvm-shutdown --all --wait $ systemctl stop qubesd $ sed 's/ driver="file"/ driver="file-reflink"/g' -i.bak /var/lib/qubes/qubes.xml $ systemctl start qubesd $ sudo rm -f /path/to/pool/*/*/*-cow.img* If the user tries to create a fresh file-reflink pool on a filesystem that doesn't support reflinks, qvm-pool will abort and mention the 'setup_check=no' option. Which can be passed to force a fallback on regular sparse copies, with of course lots of time/space overhead. The same fallback code is also used when initially cloning a VM from a foreign pool, or from another file-reflink pool on a different mountpoint. 'journalctl -fu qubesd' will show all file-reflink copy/rename/remove operations on VM creation/startup/shutdown/etc.
2018-02-12 22:20:05 +01:00
default three drivers are provided: :py:class:`qubes.storage.file.FilePool`
(named `file`), :py:class:`qubes.storage.reflink.ReflinkPool` (named
`file-reflink`), and :py:class:`qubes.storage.lvm.ThinPool` (named `lvm_thin`).
But the API allow to implement variety of other drivers (like additionally
encrypted storage, external disk, drivers using special features of some
file-reflink, a storage driver optimized for CoW filesystems This adds the file-reflink storage driver. It is never selected automatically for pool creation, especially not the creation of 'varlibqubes' (though it can be used if set up manually). The code is quite small: reflink.py lvm.py file.py + block-snapshot sloccount 334 lines 447 (134%) 570 (171%) Background: btrfs and XFS (but not yet ZFS) support instant copies of individual files through the 'FICLONE' ioctl behind 'cp --reflink'. Which file-reflink uses to snapshot VM image files without an extra device-mapper layer. All the snapshots are essentially freestanding; there's no functional origin vs. snapshot distinction. In contrast to 'file'-on-btrfs, file-reflink inherently avoids CoW-on-CoW. Which is a bigger issue now on R4.0, where even AppVMs' private volumes are CoW. (And turning off the lower, filesystem-level CoW for 'file'-on-btrfs images would turn off data checksums too, i.e. protection against bit rot.) Also in contrast to 'file', all storage features are supported, including - any number of revisions_to_keep - volume.revert() - volume.is_outdated - online fstrim/discard Example tree of a file-reflink pool - *-dirty.img are connected to Xen: - /var/lib/testpool/appvms/foo/volatile-dirty.img - /var/lib/testpool/appvms/foo/root-dirty.img - /var/lib/testpool/appvms/foo/root.img - /var/lib/testpool/appvms/foo/private-dirty.img - /var/lib/testpool/appvms/foo/private.img - /var/lib/testpool/appvms/foo/private.img@2018-01-02T03:04:05Z - /var/lib/testpool/appvms/foo/private.img@2018-01-02T04:05:06Z - /var/lib/testpool/appvms/foo/private.img@2018-01-02T05:06:07Z - /var/lib/testpool/appvms/bar/... - /var/lib/testpool/appvms/... - /var/lib/testpool/template-vms/fedora-26/... - /var/lib/testpool/template-vms/... It looks similar to a 'file' pool tree, and in fact file-reflink is drop-in compatible: $ qvm-shutdown --all --wait $ systemctl stop qubesd $ sed 's/ driver="file"/ driver="file-reflink"/g' -i.bak /var/lib/qubes/qubes.xml $ systemctl start qubesd $ sudo rm -f /path/to/pool/*/*/*-cow.img* If the user tries to create a fresh file-reflink pool on a filesystem that doesn't support reflinks, qvm-pool will abort and mention the 'setup_check=no' option. Which can be passed to force a fallback on regular sparse copies, with of course lots of time/space overhead. The same fallback code is also used when initially cloning a VM from a foreign pool, or from another file-reflink pool on a different mountpoint. 'journalctl -fu qubesd' will show all file-reflink copy/rename/remove operations on VM creation/startup/shutdown/etc.
2018-02-12 22:20:05 +01:00
filesystems, etc).
Most of storage API focus on storage volumes. Each volume have at least those
properties:
- :py:attr:`~qubes.storage.Volume.rw` - should the volume be available
read-only or read-write to the domain
- :py:attr:`~qubes.storage.Volume.snap_on_start` - should the domain start
with its own state of the volume, or rather a snapshot of its template volume
(pointed by a :py:attr:`~qubes.storage.Volume.source` property). This can be
set to `True` only if a domain do have `template` property (AppVM and DispVM).
If the domain's template is running already, the snapshot should be made out of
the template's before its startup.
- :py:attr:`~qubes.storage.Volume.save_on_stop` - should the volume state be
saved or discarded on domain
stop. In either case, while the domain is running, volume's current state
should not be committed immediately. This is to allow creating snapshots of the
volume's state from before domain start (see
:py:attr:`~qubes.storage.Volume.snap_on_start`).
- :py:attr:`~qubes.storage.Volume.revisions_to_keep` - number of volume
revisions to keep. If greater than zero, at each domain stop (and if
:py:attr:`~qubes.storage.Volume.save_on_stop` is `True`) new revision is saved
and old ones exceeding :py:attr:`~qubes.storage.Volume.revisions_to_keep` limit
are removed.
- :py:attr:`~qubes.storage.Volume.source` - source volume for
:py:attr:`~qubes.storage.Volume.snap_on_start` volumes
- :py:attr:`~qubes.storage.Volume.vid` - pool specific volume identifier, must
be unique inside given pool
- :py:attr:`~qubes.storage.Volume.pool` - storage pool object owning this volume
- :py:attr:`~qubes.storage.Volume.name` - name of the volume inside owning
domain (like `root`, or `private`)
- :py:attr:`~qubes.storage.Volume.size` - size of the volume, in bytes
Storage pool driver may define additional properties.
Storage pool driver API
-----------------------
Storage pool driver need to implement two classes:
- pool class - inheriting from :py:class:`qubes.storage.Pool`
- volume class - inheriting from :py:class:`qubes.storage.Volume`
Pool class should be registered with `qubes.storage` entry_point, under the
name of storage pool driver. Volume class instances should be returned by
:py:meth:`qubes.storage.Pool.init_volume` method of pool class instance.
Methods required to be implemented by the pool class:
- :py:meth:`~qubes.storage.Pool.init_volume` - return instance of appropriate
volume class; this method should not alter any persistent disk state, it is
used to instantiate both existing volumes and create new ones
- :py:meth:`~qubes.storage.Pool.setup` - setup new storage pool
- :py:meth:`~qubes.storage.Pool.destroy` - destroy storage pool
Methods and properties required to be implemented by the volume class:
- :py:meth:`~qubes.storage.Volume.create` - create volume on disk
- :py:meth:`~qubes.storage.Volume.remove` - remove volume from disk
- :py:meth:`~qubes.storage.Volume.start` - prepare the volume for domain start;
this include making a snapshot if
:py:attr:`~qubes.storage.Volume.snap_on_start` is `True`
- :py:meth:`~qubes.storage.Volume.stop` - cleanup after domain shutdown; this
include committing changes to the volume if
:py:attr:`~qubes.storage.Volume.save_on_stop` is `True`
- :py:meth:`~qubes.storage.Volume.export` - return a path to be read to extract
volume data; for complex formats, this can be a pipe (connected to some
data-extracting process)
- :py:meth:`~qubes.storage.Volume.export_end` - cleanup after exporting the
data; this function is called when the path returned by
:py:meth:`~qubes.storage.Volume.export` is not used anymore. This method
optional - some storage drivers may not implement it if not needed.
- :py:meth:`~qubes.storage.Volume.import_data` - return a path the data should
be written to, to import volume data; for complex formats, this can be pipe
(connected to some data-importing process)
- :py:meth:`~qubes.storage.Volume.import_data_end` - finish data import
operation (cleanup temporary files etc); this methods is called always after
:py:meth:`~qubes.storage.Volume.import_data` regardless if operation was
successful or not
- :py:meth:`~qubes.storage.Volume.import_volume` - import data from another volume
- :py:meth:`~qubes.storage.Volume.resize` - resize volume
- :py:meth:`~qubes.storage.Volume.revert` - revert volume state to a given revision
- :py:attr:`~qubes.storage.Volume.revisions` - collection of volume revisions (to use
with :py:meth:`qubes.storage.Volume.revert`)
- :py:meth:`~qubes.storage.Volume.is_dirty` - is volume properly committed
after domain shutdown? Applies only to volumes with
:py:attr:`~qubes.storage.Volume.save_on_stop` set to `True`
- :py:meth:`~qubes.storage.Volume.is_outdated` - have the source volume started
since domain startup? applies only to volumes with
:py:attr:`~qubes.storage.Volume.snap_on_start` set to `True`
- :py:attr:`~qubes.storage.Volume.config` - volume configuration, this should
be enough to later reinstantiate the same volume object
- :py:meth:`~qubes.storage.Volume.block_device` - return
:py:class:`qubes.storage.BlockDevice` instance required to configure volume in
libvirt
Some storage pool drivers can provide limited functionality only - for example
support only `volatile` volumes (those with
:py:attr:`~qubes.storage.Volume.snap_on_start` is `False`,
:py:attr:`~qubes.storage.Volume.save_on_stop` is `False`, and
:py:attr:`~qubes.storage.Volume.rw` is `True`). In that case, it should raise
:py:exc:`NotImplementedError` in :py:meth:`qubes.storage.Pool.init_volume` when
trying to instantiate unsupported volume.
Note that pool driver should be prepared to recover from power loss before
stopping a domain - so, if volume have
:py:attr:`~qubes.storage.Volume.save_on_stop` is `True`, and
:py:meth:`qubes.storage.Volume.stop` wasn't called, next
:py:meth:`~qubes.storage.Volume.start` should pick up previous (not committed)
state.
See specific methods documentation for details.
Module contents
---------------
.. automodule:: qubes.storage
:members:
:show-inheritance:
.. vim: ts=3 sw=3 et