guestfs-faq - libguestfs Frequently Asked Questions (FAQ)
libguestfs is a way to create, access and modify disk images. You can look inside disk images, modify the files they contain, create them from scratch, resize them, and much more. It's especially useful from scripts and programs and from the command line.
libguestfs is a C library (hence "lib-"), and a set of tools built on this library, and bindings for many common programming languages.
For more information about what libguestfs can do read the introduction on the home page (http://libguestfs.org).
Virt tools (website: http://virt-tools.org) are a whole set of virtualization management tools aimed at system administrators. Some of them come from libguestfs, some from libvirt and many others from other open source projects. So virt tools is a superset of libguestfs. However libguestfs comes with many important tools. See http://libguestfs.org for a full list.
No!
libvirt is not a requirement for libguestfs.
libguestfs works with any disk image, including ones created in VMware, KVM, qemu, VirtualBox, Xen, and many other hypervisors, and ones which you have created from scratch.
Red Hat sponsors (ie. pays for) development of libguestfs and a huge number of other open source projects. But you can run libguestfs and the virt tools on many different Linux distros and Mac OS X. We try our best to support all Linux distros as first-class citizens. Some virt tools have been ported to Windows.
- vs. kpartx
-
Libguestfs takes a different approach from kpartx. kpartx needs root, and mounts filesystems on the host kernel (which can be insecure - see "SECURITY" in guestfs(3)). Libguestfs isolates your host kernel from guests, is more flexible, scriptable, supports LVM, doesn't require root, is isolated from other processes, and cleans up after itself. Libguestfs is more than just file access because you can use it to create images from scratch.
- vs. vdfuse
-
vdfuse is like kpartx but for VirtualBox images. See the kpartx comparison above. You can use libguestfs on the partition files exposed by vdfuse, although it's not necessary since libguestfs can access VirtualBox images directly.
- vs. qemu-nbd
-
NBD (Network Block Device) is a protocol for exporting block devices over the network. qemu-nbd is an NBD server which can handle any disk format supported by qemu (eg. raw, qcow2). You can use libguestfs and qemu-nbd or nbdkit together to access block devices over the network, for example:
guestfish -a nbd://remote - vs. mounting filesystems in the host
-
Mounting guest filesystems in the host is insecure and should be avoided completely for untrusted guests. Use libguestfs to provide a layer of protection against filesystem exploits. See also guestmount(1).
- vs. parted
-
Libguestfs supports LVM. Libguestfs uses parted and provides most parted features through the libguestfs API.
The simplest method is:
guestfish --versionLibguestfs development happens along an unstable branch and we periodically create a stable branch which we backport stable patches to. To find out more, read "LIBGUESTFS VERSION NUMBERS" in guestfs(3).
If you are a Red Hat customer using Red Hat Enterprise Linux, please contact Red Hat Support: http://redhat.com/support
There is a mailing list, mainly for development, but users are also welcome to ask questions about libguestfs and the virt tools: https://www.redhat.com/mailman/listinfo/libguestfs
You can also talk to us on IRC channel #libguestfs on FreeNode. We're not always around, so please stay in the channel after asking your question and someone will get back to you.
For other virt tools (not ones supplied with libguestfs) there is a general virt tools mailing list: https://www.redhat.com/mailman/listinfo/virt-tools-list
Please use the following link to enter a bug in Bugzilla:
https://bugzilla.redhat.com/enter_bug.cgi?component=libguestfs&product=Virtualization+Tools
Include as much detail as you can and a way to reproduce the problem.
Include the full output of libguestfs-test-tool(1).
See also "LIBGUESTFS GOTCHAS" in guestfs(3) for some "gotchas" with using the libguestfs API.
This obscure error is in fact an SELinux failure. You have to enable the following SELinux boolean:
setsebool -P virt_use_execmem=onFor more information see https://bugzilla.redhat.com/show_bug.cgi?id=806106.
[This error message was changed in libguestfs 1.21.18 to something more explanatory.]
This error indicates that qemu failed or the host kernel could not boot. To get further information about the failure, you have to run:
libguestfs-test-toolIf, after using this, you still don't understand the failure, contact us (see previous section).
libguestfs: error: cannot find any suitable libguestfs supermin, fixed or old-style appliance on LIBGUESTFS_PATH
If you see any of these errors on Debian/Ubuntu, you need to run the following command:
sudo update-guestfs-applianceYou get a permission denied error when opening a disk image, even though you are running libguestfs as root.
This is caused by libvirt, and so only happens when using the libvirt backend. When run as root, libvirt decides to run the qemu appliance as user qemu.qemu. Unfortunately this usually means that qemu cannot open disk images, especially if those disk images are owned by root, or are present in directories which require root access.
There is a bug open against libvirt to fix this: https://bugzilla.redhat.com/show_bug.cgi?id=1045069
You can work around this by one of the following methods:
Switch to the direct backend:
export LIBGUESTFS_BACKEND=directDon't run libguestfs as root.
Chmod the disk image and any parent directories so that the qemu user can access them.
(Nasty) Edit
/etc/libvirt/qemu.confand change theusersetting.
Note: If this error happens when you are using a distro package of libguestfs (eg. from Fedora, Debian, etc) then file a bug against the distro. This is not an error which normal users should ever see if the distro package has been prepared correctly.
This error happens during the supermin boot phase of starting the appliance:
supermin: mounting new root on /root
supermin: chroot
execl: /init: Permission denied
supermin: debug: listing directory /
[...followed by a lot of debug output...]This is a complicated bug related to supermin(1) appliances. The appliance is constructed by copying files like /bin/bash and many libraries from the host. The file hostfiles lists the files that should be copied from the host into the appliance. If some files don't exist on the host then they are missed out, but if these files are needed in order to (eg) run /bin/bash then you'll see the above error.
Diagnosing the problem involves studying the libraries needed by /bin/bash, ie:
ldd /bin/bashcomparing that with hostfiles, with the files actually available in the host filesystem, and with the debug output printed in the error message. Once you've worked out which file is missing, install that file using your package manager and try again.
You should also check that files like /init and /bin/bash (in the appliance) are executable. The debug output shows file modes.
- Fedora ≥ 11
-
Use:
yum install '*guestf*'For the latest builds, see: http://koji.fedoraproject.org/koji/packageinfo?packageID=8391
- Red Hat Enterprise Linux
-
- RHEL 5
-
The version shipped in official RHEL 5 is very old and should not be used except in conjunction with virt-v2v. Use the up-to-date libguestfs 1.20 package in EPEL 5: https://fedoraproject.org/wiki/EPEL
- RHEL 6
-
It is part of the default install. On RHEL 6 and 7 (only) you have to install
libguestfs-winsupportto get Windows guest support. - RHEL 7
-
It will be part of the default install, and based on libguestfs 1.22. You will need to install
libguestfs-winsupportseparately to get Windows guest support.
- Debian and Ubuntu
-
After installing libguestfs you need to do:
sudo update-guestfs-applianceOn Ubuntu only:
sudo chmod 0644 /boot/vmlinuz*You may need to add yourself to the
kvmgroup:sudo usermod -a -G kvm yourlogin- Debian Squeeze (6)
-
Hilko Bengen has built libguestfs in squeeze backports: http://packages.debian.org/search?keywords=guestfs&searchon=names§ion=all&suite=squeeze-backports
- Debian Wheezy and later (7+)
-
Hilko Bengen supports libguestfs on Debian. Official Debian packages are available: http://packages.debian.org/search?keywords=libguestfs
- Ubuntu
-
We don't have a full time Ubuntu maintainer, and the packages supplied by Canonical (which are outside our control) are sometimes broken.
Canonical decided to change the permissions on the kernel so that it's not readable except by root. This is completely stupid, but they won't change it (https://bugs.launchpad.net/ubuntu/+source/linux/+bug/759725). So every user should do this:
sudo chmod 0644 /boot/vmlinuz*- Ubuntu 12.04
-
libguestfs in this version of Ubuntu works, but you need to update febootstrap and seabios to the latest versions.
You need febootstrap ≥ 3.14-2 from: http://packages.ubuntu.com/precise/febootstrap
After installing or updating febootstrap, rebuild the appliance:
sudo update-guestfs-applianceYou need seabios ≥ 0.6.2-0ubuntu2.1 or ≥ 0.6.2-0ubuntu3 from: http://packages.ubuntu.com/precise-updates/seabios or http://packages.ubuntu.com/quantal/seabios
Also you need to do (see above):
sudo chmod 0644 /boot/vmlinuz*
- Gentoo
-
Libguestfs was added to Gentoo in 2012-07 by Andreis Vinogradovs (libguestfs) and Maxim Koltsov (mainly hivex). Do:
emerge libguestfs - SuSE
-
Libguestfs was added to SuSE in 2012 by Olaf Hering.
- ArchLinux
-
Libguestfs was added to the AUR in 2010.
- Other Linux distro
-
Compile from source (next section).
- Other non-Linux distro
-
You'll have to compile from source, and port it.
If your Linux distro has a working port of supermin (that is, Fedora, Red Hat Enterprise Linux ≥ 6.3, Debian, Ubuntu and ArchLinux) then you should just be able to compile from source in the usual way. Download the latest tarball from http://libguestfs.org/download, unpack it, and start by reading the README file.
If you don't have supermin, you will need to use the "fixed appliance method". See: http://libguestfs.org/download/binaries/appliance/
Patches to port supermin to more Linux distros are welcome.
Note: This issue is fixed in Fedora ≥ 19 and RHEL ≥ 6.5.
Because of the complexity of building the libguestfs appliance, the source RPMs provided cannot be rebuilt directly using rpmbuild or mock.
If you use Koji (which is open source software and may be installed locally), then the SRPMs can be rebuilt in Koji. https://fedoraproject.org/wiki/Koji
If you don't have or want to use Koji, then you have to give libguestfs access to the network so it can download the RPMs for building the appliance. You also need to set an RPM macro to tell libguestfs to use the network. Put the following line into a file called $HOME/.rpmmacros:
%libguestfs_buildnet 1If you are using mock, do:
mock -D '%libguestfs_buildnet 1' [etc]Note for Fedora/RHEL users: This configuration is the default starting with Fedora 18 and RHEL 7. If you find any problems, please let us know or file a bug.
SVirt provides a hardened appliance using SELinux, making it very hard for a rogue disk image to "escape" from the confinement of libguestfs and damage the host (it's fair to say that even in standard libguestfs this would be hard, but sVirt provides an extra layer of protection for the host and more importantly protects virtual machines on the same host from each other).
Currently to enable sVirt you will need libvirt ≥ 0.10.2 (1.0 or later preferred), libguestfs ≥ 1.20, and the SELinux policies from recent Fedora. If you are not running Fedora 18+, you will need to make changes to your SELinux policy - contact us on the mailing list.
Once you have the requirements, do:
./configure --with-default-backend=libvirt # libguestfs >= 1.22
./configure --with-default-attach-method=libvirt # libguestfs <= 1.20
makeSet SELinux to Enforcing mode, and sVirt should be used automatically.
All, or almost all, features of libguestfs should work under sVirt. There is one known shortcoming: virt-rescue(1) will not use libvirt (hence sVirt), but falls back to direct launch of qemu. So you won't currently get the benefit of sVirt protection when using virt-rescue.
You can check if sVirt is being used by enabling libvirtd logging (see /etc/libvirt/libvirtd.log), killing and restarting libvirtd, and checking the log files for "Setting SELinux context on ..." messages.
In theory sVirt should support AppArmor, but we have not tried it. It will almost certainly require patching libvirt and writing an AppArmor policy.
That's because it does a lot of things.
Libguestfs -- as it is packaged for Fedora -- satisfies the following conditions:
The Fedora package is full featured, that is, it supports every possible feature of libguestfs (every filesystem, every filesystem tool, etc.)
A common request is to split up libguestfs into separate feature areas so you could, say, install XFS support and NTFS support separately. This is not possible right now.
The download size of the libguestfs package is relatively small (ie. not ten's of megabytes as it would be if it included a complete, "statically linked" appliance).
The Fedora package automatically updates itself if there is a security update. It doesn't include a huge static blob that has to be rebuilt and users have to re-download if there is an update.
Able to be installed without needing direct network access. This is important when using closed networks, privately mirrored repositories or RHN Satellite.
The Fedora package can be tested during the build.
If you want to drop any one of those conditions, then you can package libguestfs differently and make it have fewer dependencies, fewer features or a faster start up time:
- 1. (full featured)
-
Take
appliance/packagelist.inin the source, and comment out any features you don't actually care about. For example if you never anticipate editing a Windows guest, remove all the ntfs-related packages. You can get away with fewer dependencies. - 2. (download size) / 3. (updates)
-
Use libguestfs-make-fixed-appliance(1) to build a compressed appliance. Bundle this with your package and set
$LIBGUESTFS_PATHto point to it. Users will have to download this large appliance, but no dependencies are needed, and supermin-helper(1) is not used. - 4. (network access)
-
Reconstruct and cache the appliance once during package install. The Debian packaging currently works like this, but requires network access during package install.
- 5. (tests)
-
Don't run any tests during the build. The build will be much faster, but also less likely to work correctly.
Note that running the tests in
tests/qemuis probably a good idea, since those are sanity tests. Also you should domake quickcheckto ensure libguestfs is basically working.
In Fedora ≥ 18 and RHEL ≥ 7, libguestfs uses libvirt to manage the appliance. Previously (and upstream) libguestfs runs qemu directly:
+----------------------------------+
| libguestfs |
+----------------+-----------------+
| direct backend | libvirt backend |
+----------------+-----------------+
| |
v v
+-------+ +----------+
| qemu | | libvirtd |
+-------+ +----------+
|
v
+-------+
| qemu |
+-------+
upstream Fedora 18+
non-Fedora RHEL 7+
non-RHELThe libvirt backend is more sophisticated, supporting SELinux/sVirt (see above), hotplugging and more. It is, however, more complex and so less robust.
If you have permissions problems using the libvirt backend, you can switch to the direct backend by setting this environment variable:
export LIBGUESTFS_BACKEND=directbefore running any libguestfs program or virt tool.
This may improve the stability and performance of libguestfs on Fedora and RHEL.
Any time after installing libguestfs, run the following commands as root:
mkdir -p /usr/local/lib/guestfs/appliance
libguestfs-make-fixed-appliance /usr/local/lib/guestfs/appliance
ls -l /usr/local/lib/guestfs/applianceNow set the following environment variable before using libguestfs or any virt tool:
export LIBGUESTFS_PATH=/usr/local/lib/guestfs/applianceOf course you can change the path to any directory you want. You can share the appliance across machines that have the same architecture (eg. all x86-64), but note that libvirt will prevent you from sharing the appliance across NFS because of permissions problems (so either switch to the direct backend or don't use NFS).
By far the most important thing you can do is to install and properly configure Squid. Note that the default configuration that ships with Squid is rubbish, so configuring it is not optional.
A very good place to start with Squid configuration is here: https://fedoraproject.org/wiki/Extras/MockTricks#Using_Squid_to_Speed_Up_Mock_package_downloads
Make sure Squid is running, and that the environment variables $http_proxy and $ftp_proxy are pointing to it.
With Squid running and correctly configured, appliance builds should be reduced to a few minutes.
Hilko Bengen suggests using "approx" which is a Debian archive proxy (http://packages.debian.org/approx). This tool is documented on Debian in the approx(8) manual page.
Note: Most of the information in this section has moved: guestfs-performance(1).
If the underlying disk is not fully allocated (eg. sparse raw or qcow2) then writes can be slow because the host operating system has to do costly disk allocations while you are writing. The solution is to use a fully allocated format instead, ie. non-sparse raw, or qcow2 with the preallocation=metadata option.
libguestfs caches a large-ish appliance in:
/var/tmp/.guestfs-<UID>If the environment variable TMPDIR is defined, then $TMPDIR/.guestfs-<UID> is used instead.
It is safe to delete this directory when you are not using libguestfs.
If the input to virt-sparsify(1) is raw, then the output will be raw sparse. Make sure you are measuring the output with a tool which understands sparseness such as du -sh. It can make a huge difference:
$ ls -lh test1.img
-rw-rw-r--. 1 rjones rjones 100M Aug 8 08:08 test1.img
$ du -sh test1.img
3.6M test1.img(Compare the apparent size 100M vs the actual size 3.6M)
If all this confuses you, use a non-sparse output format by specifying the --convert option, eg:
virt-sparsify --convert qcow2 disk.raw disk.qcow2Resizing a disk image is very tricky -- especially making sure that you don't lose data or break the bootloader. The current method effectively creates a new disk image and copies the data plus bootloader from the old one. If something goes wrong, you can always go back to the original.
If we were to make virt-resize work in-place then there would have to be limitations: for example, you wouldn't be allowed to move existing partitions (because moving data across the same disk is most likely to corrupt data in the event of a power failure or crash), and LVM would be very difficult to support (because of the almost arbitrary mapping between LV content and underlying disk blocks).
Another method we have considered is to place a snapshot over the original disk image, so that the original data is untouched and only differences are recorded in the snapshot. You can do this today using qemu-img create + virt-resize, but qemu currently isn't smart enough to recognize when the same block is written back to the snapshot as already exists in the backing disk, so you will find that this doesn't save you any space or time.
In summary, this is a hard problem, and what we have now mostly works so we are reluctant to change it.
Eventually we plan to make virt-sparsify work on disk images in-place, instead of copying the disk image. However it requires several changes to both the Linux kernel and qemu which are slowly making their way upstream (thanks to the tireless efforts of Paolo Bonzini). Then we will have to modify virt-sparsify to support this. Finally there will be some integration work required to make sure all the pieces work together.
Even with this implemented there may be some limitations: For example, it requires completely different steps (and is probably harder) to sparsify a disk image that is stored on a SAN LUN, compared to one which is stored in a local raw image file, so you can expect that different storage and backing formats will become supported at different times. Some backing filesystems / formats may never support sparsification (eg. disk images stored on VFAT, old-style non-thin LVs).
Opening remote libvirt guests is not supported at this time. For example this won't work:
guestfish -c qemu://remote/system -d GuestTo open remote disks you have to export them somehow, then connect to the export. For example if you decided to use NBD:
remote$ qemu-nbd -t -p 10809 guest.img
local$ guestfish -a nbd://remote:10809 -iOther possibilities include ssh (if qemu is recent enough), NFS or iSCSI. See "REMOTE STORAGE" in guestfs(3).
You have a disk image located inside another system that requires access via a library / HTTP / REST / proprietary API, or is compressed or archived in some way. (One example would be remote access to OpenStack glance images without actually downloading them.)
We have a sister project called nbdkit (https://github.com/libguestfs/nbdkit). This project lets you turn any disk source into an NBD server. Libguestfs can access NBD servers directly, eg:
guestfish -a nbd://remotenbdkit is liberally licensed, so you can link it to or include it in proprietary libraries and code. It also has a simple, stable plugin API so you can easily write plugins against the API which will continue to work in future.
Error opening VMDK disks: "uses a vmdk feature which is not supported by this qemu version: VMDK version 3"
Qemu (and hence libguestfs) only supports certain VMDK disk images. Others won't work, giving this or similar errors.
Ideally someone would fix qemu to support the latest VMDK features, but in the meantime you have three options:
If the guest is hosted on a live, reachable ESX server, then locate and download the disk image called somename
-flat.vmdk. Despite the name, this is a raw disk image, and can be opened by anything.If you have a recent enough version of qemu and libguestfs, then you may be able to access this disk image remotely using either HTTPS or ssh. See "REMOTE STORAGE" in guestfs(3).
Use VMware's proprietary vdiskmanager tool to convert the image to raw format.
Use nbdkit with the proprietary VDDK plugin to live export the disk image as an NBD source. This should allow you to read and write the VMDK file.
The UFS filesystem format has many variants, and these are not self-identifying. The Linux kernel has to be told which variant of UFS it has to use, which libguestfs cannot know.
You have to pass the right ufstype mount option when mounting these filesystems.
See https://www.kernel.org/doc/Documentation/filesystems/ufs.txt
Windows ReFS is Microsoft's ZFS/Btrfs copy. This filesystem has not yet been reverse engineered and implemented in the Linux kernel, and therefore libguestfs doesn't support it. At the moment it seems to be very rare "in the wild".
Typical symptoms of this problem:
You get an error when you create a file where the filename contains non-ASCII characters, particularly non 8-bit characters from Asian languages (Chinese, Japanese, etc). The filesystem is VFAT.
When you list a directory from a VFAT filesystem, filenames appear as question marks.
This is a design flaw of the GNU/Linux system.
VFAT stores long filenames as UTF-16 characters. When opening or returning filenames, the Linux kernel has to translate these to some form of 8 bit string. UTF-8 would be the obvious choice, except for Linux users who persist in using non-UTF-8 locales (the user's locale is not known to the kernel because it's a function of libc).
Therefore you have to tell the kernel what translation you want done when you mount the filesystem. The two methods are the iocharset parameter (which is not relevant to libguestfs) and the utf8 flag.
So to use a VFAT filesystem you must add the utf8 flag when mounting. From guestfish, use:
><fs> mount-options utf8 /dev/sda1 /or on the guestfish command line:
guestfish [...] -m /dev/sda1:/:utf8or from the API:
guestfs_mount_options (g, "utf8", "/dev/sda1", "/");The kernel will then translate filenames to and from UTF-8 strings.
We considered adding this mount option transparently, but unfortunately there are several problems with doing that:
On some Linux systems, the
utf8mount option doesn't work. We don't precisely understand what systems or why, but this was reliably reported by one user.It would prevent you from using the
iocharsetparameter because it is incompatible withutf8. It is probably not a good idea to use this parameter, but we don't want to prevent it.
The filesystem was not prepared correctly with mkisofs or genisoimage. Make sure the filesystem was created using Joliet and/or Rock Ridge extensions. libguestfs does not require any special mount options to handle the filesystem.
We recommend you start by reading the API overview: "API OVERVIEW" in guestfs(3).
Although the API overview covers the C API, it is still worth reading even if you are going to use another programming language, because the API is the same, just with simple logical changes to the names of the calls:
C guestfs_ln_sf (g, target, linkname);
Python g.ln_sf (target, linkname);
OCaml g#ln_sf target linkname;
Perl $g->ln_sf (target, linkname);
Shell (guestfish) ln-sf target linkname
PHP guestfs_ln_sf ($g, $target, $linkname);Once you're familiar with the API overview, you should look at this list of starting points for other language bindings: "USING LIBGUESTFS WITH OTHER PROGRAMMING LANGUAGES" in guestfs(3).
In general, yes. However this is not legal advice - read the license that comes with libguestfs, and if you have specific questions contact a lawyer.
In the source tree the license is in the file COPYING.LIB (LGPLv2+ for the library and bindings) and COPYING (GPLv2+ for the standalone programs).
Please supply all the information in this checklist, in an email sent to libguestfs @ redhat.com:
What are you trying to achieve?
What exact commands did you run?
What was the precise error / output of these commands?
Enable debugging, run the commands again, and capture the complete output. Do not edit the output.
export LIBGUESTFS_DEBUG=1 export LIBGUESTFS_TRACE=1Include the version of libguestfs, the operating system version, and how you installed libguestfs (eg. from source,
yum install, etc.)If no libguestfs program seems to work at all, run the program below and paste the complete, unedited output into the email:
libguestfs-test-tool
There are two LIBGUESTFS_* environment variables you can set in order to get more information from libguestfs.
LIBGUESTFS_TRACE-
Set this to 1 and libguestfs will print out each command / API call in a format which is similar to guestfish commands.
LIBGUESTFS_DEBUG-
Set this to 1 in order to enable massive amounts of debug messages. If you think there is some problem inside the libguestfs appliance, then you should use this option.
To set these from the shell, do this before running the program:
export LIBGUESTFS_TRACE=1
export LIBGUESTFS_DEBUG=1For csh/tcsh the equivalent commands would be:
setenv LIBGUESTFS_TRACE 1
setenv LIBGUESTFS_DEBUG 1For further information, see: "ENVIRONMENT VARIABLES" in guestfs(3).
You can use the same environment variables above. Alternatively use the guestfish options -x (to trace commands) or -v (to get the full debug output), or both.
For further information, see: guestfish(1).
Call "guestfs_set_trace" in guestfs(3) to enable command traces, and/or "guestfs_set_verbose" in guestfs(3) to enable debug messages.
For best results, call these functions as early as possible, just after creating the guestfs handle if you can, and definitely before calling launch.
Use the event API. For examples, see: "SETTING CALLBACKS TO HANDLE EVENTS" in guestfs(3) and the examples/debug-logging.c program in the libguestfs sources.
Enable debugging and then read this documentation on the appliance boot process: "INTERNALS" in guestfs(3).
Enable debugging and look at the full output. If you cannot work out what is going on, file a bug report, including the complete output of libguestfs-test-tool(1).
We offer a command called guestmount(1) which lets you mount guest filesystems on the host. This is implemented as a FUSE module. Why don't we just implement the whole of libguestfs using this mechanism, instead of having the large and rather complicated API?
The reasons are twofold. Firstly, libguestfs offers API calls for doing things like creating and deleting partitions and logical volumes, which don't fit into a filesystem model very easily. Or rather, you could fit them in: for example, creating a partition could be mapped to mkdir /fs/hda1 but then you'd have to specify some method to choose the size of the partition (maybe echo 100M > /fs/hda1/.size), and the partition type, start and end sectors etc., but once you've done that the filesystem-based API starts to look more complicated than the call-based API we currently have.
The second reason is for efficiency. FUSE itself is reasonably efficient, but it does make lots of small, independent calls into the FUSE module. In guestmount these have to be translated into messages to the libguestfs appliance which has a big overhead (in time and round trips). For example, reading a file in 64 KB chunks is inefficient because each chunk would turn into a single round trip. In the libguestfs API it is much more efficient to download an entire file or directory through one of the streaming calls like guestfs_download or guestfs_tar_out.
The problems are similar to the problems with FUSE.
GVFS is a better abstraction than POSIX/FUSE. There is an FTP backend for GVFS, which is encouraging because FTP is conceptually similar to the libguestfs API. However the GVFS FTP backend makes multiple simultaneous connections in order to keep interactivity, which we can't easily do with libguestfs.
When you add a disk read-only, libguestfs places a writable overlay on top of the underlying disk. Writes go into this overlay, and are discarded when the handle is closed (or guestfish etc. exits).
There are two reasons for doing it this way: Firstly read-only disks aren't possible in many cases (eg. IDE simply doesn't support them, so you couldn't have an IDE-emulated read-only disk, although this is not common in real libguestfs installations).
Secondly and more importantly, even if read-only disks were possible, you wouldn't want them. Mounting any filesystem that has a journal, even mount -o ro, causes writes to the filesystem because the journal has to be replayed and metadata updated. If the disk was truly read-only, you wouldn't be able to mount a dirty filesystem.
To make it usable, we create the overlay as a place to temporarily store these writes, and then we discard it afterwards. This ensures that the underlying disk is always untouched.
Note also that there is a regression test for this when building libguestfs (in tests/qemu). This is one reason why it's important for packagers to run the test suite.
No! The --ro option only affects disks added on the command line, ie. using -a and -d options.
In guestfish, if you use the add command, then disk is added read-write (unless you specify the readonly:true flag explicitly with the command).
Usually this is not a good idea. The question is answered in more detail in this mailing list posting: https://www.redhat.com/archives/libguestfs/2010-August/msg00024.html
See also the next question.
This command will usually not work:
guestfish --ro -a /dev/vg/my_root_fs run : fsck /dev/sdaThe reason for this is that qemu creates a snapshot over the original filesystem, but it doesn't create a strict point-in-time snapshot. Blocks of data on the underlying filesystem are read by qemu at different times as the fsck operation progresses, with host writes in between. The result is that fsck sees massive corruption (imaginary, not real!) and fails.
What you have to do is to create a point-in-time snapshot. If it's a logical volume, use an LVM2 snapshot. If the filesystem is located inside something like a btrfs/ZFS file, use a btrfs/ZFS snapshot, and then run the fsck on the snapshot. In practice you don't need to use libguestfs for this -- just run /sbin/fsck directly.
Creating point-in-time snapshots of host devices and files is outside the scope of libguestfs, although libguestfs can operate on them once they are created.
A lot of people are confused by the two superficially similar tools we provide:
$ guestfish --ro -a guest.img
><fs> run
><fs> fsck /dev/sda1
$ virt-rescue --ro guest.img
><rescue> /sbin/fsck /dev/sda1And the related question which then arises is why you can't type in full shell commands with all the --options in guestfish (but you can in virt-rescue(1)).
guestfish(1) is a program providing structured access to the guestfs(3) API. It happens to be a nice interactive shell too, but its primary purpose is structured access from shell scripts. Think of it more like a language binding, like Python and other bindings, but for shell. The key differentiating factor of guestfish (and the libguestfs API in general) is the ability to automate changes.
virt-rescue(1) is a free-for-all freeform way to boot the libguestfs appliance and make arbitrary changes to your VM. It's not structured, you can't automate it, but for making quick ad-hoc fixes to your guests, it can be quite useful.
But, libguestfs also has a "backdoor" into the appliance allowing you to send arbitrary shell commands. It's not as flexible as virt-rescue, because you can't interact with the shell commands, but here it is anyway:
><fs> debug sh "cmd arg1 arg2 ..."Note that you should not rely on this. It could be removed or changed in future. If your program needs some operation, please add it to the libguestfs API instead.
These questions are all related at a fundamental level which may not be immediately obvious.
At the guestfs(3) API level, a "disk image" is just a pile of partitions and filesystems.
In contrast, when the virtual machine boots, it mounts those filesystems into a consistent hierarchy such as:
/ (/dev/sda2)
|
+-- /boot (/dev/sda1)
|
+-- /home (/dev/vg_external/Homes)
|
+-- /usr (/dev/vg_os/lv_usr)
|
+-- /var (/dev/vg_os/lv_var)(or drive letters on Windows).
The API first of all sees the disk image at the "pile of filesystems" level. But it also has a way to inspect the disk image to see if it contains an operating system, and how the disks are mounted when the operating system boots: "INSPECTION" in guestfs(3).
Users expect some tools (like virt-cat(1)) to work with VM paths:
virt-cat fedora.img /var/log/messagesHow does virt-cat know that /var is a separate partition? The trick is that virt-cat performs inspection on the disk image, and uses that to translate the path correctly.
Some tools (including virt-cat(1), virt-edit(1), virt-ls(1)) use inspection to map VM paths. Other tools, such as virt-df(1) and virt-filesystems(1) operate entirely at the raw "big pile of filesystems" level of the libguestfs API, and don't use inspection.
guestfish(1) is in an interesting middle ground. If you use the -a and -m command line options, then you have to tell guestfish exactly how to add disk images and where to mount partitions. This is the raw API level.
If you use the -i option, libguestfs performs inspection and mounts the filesystems for you.
The error no root device found in this operating system image is related to this. It means inspection was unable to locate an operating system within the disk image you gave it. You might see this from programs like virt-cat if you try to run them on something which is just a disk image, not a virtual machine disk image.
There are some functions which are used for debugging and internal purposes which are not part of the stable API.
The debug* (or guestfs_debug*) functions, primarily "guestfs_debug" in guestfs(3) and a handful of others, are used for debugging libguestfs. Although they are not part of the stable API and thus may change or be removed at any time, some programs may want to call these while waiting for features to be added to libguestfs.
The internal-* (or guestfs_internal_*) functions are purely to be used by libguestfs itself. There is no reason for programs to call them, and programs should not try to use them. Using them will often cause bad things to happen, as well as not being part of the documented stable API.
Please send patches to the libguestfs mailing list https://www.redhat.com/mailman/listinfo/libguestfs. You don't have to be subscribed, but there will be a delay until your posting is manually approved.
Please don't use github pull requests - they will be ignored. The reasons are (a) we want to discuss and dissect patches on the mailing list, and (b) github pull requests turn into merge commits but we prefer to have a linear history.
Large new features that you intend to contribute should be discussed on the mailing list first (https://www.redhat.com/mailman/listinfo/libguestfs). This avoids disappointment and wasted work if we don't think the feature would fit into the libguestfs project.
If you want to suggest a useful feature but don't want to write the code, you can file a bug (see "GETTING HELP AND REPORTING BUGS") with "RFE: " at the beginning of the Summary line.
About 5 people have commit access to github. Patches should be posted on the list first and ACKed. The policy for ACKing and pushing patches is outlined here:
https://www.redhat.com/archives/libguestfs/2012-January/msg00023.html
Of course you can. Git makes it easy to fork libguestfs. Github makes it even easier. It's nice if you tell us on the mailing list about forks and the reasons for them.
guestfish(1), guestfs(3), http://libguestfs.org/.
Richard W.M. Jones (rjones at redhat dot com)
Copyright (C) 2012-2014 Red Hat Inc.