Skip setting various additional connection params if the transport for
the libvirt connection is not ssh based as these will be ignored and may
cause confusion as to why they do not apply.
Switch to only picking up LIBVIRT_DEFAULT_URI if no settings that could
affect the uri have been explicitly set. Some of these may not actually
appear in the URI depending on what is actually set or not set, however
it is important that should only use the env variable if the user has
not explicitly configured options relevant and instead allow them to
configure as needed.
One can now set `LIBVIRT_DEFAULT_URI` to something like
`qemu+ssh://user@beefy/system` along with `connect_via_ssh` option to
true and use vagrant on remote machines, without needing to
duplicate/hardcode remote server username and identity key inside the
Vagrantfile.
Partial Fixes: #1217
Solves vagrant not detecting end of ssh connection when a proxy is
used, described in #921
Allows ssh argument population to append arguments only when
respective values are not nil, avoiding creation of invalid
proxy_command when ssh-agent or ssh config based connections are used
Refactor WaitTillUp action to make use of the domain IP address
retrieval code in the driver to ensure a single place to maintain.
Remove references to machine option for driver where already should be
available as an instance variable.
Adds disk driver options: io, copy_on_read, discard and detect_zeroes
for both the primary vm disk and additional disks.
Minor refactor of existing volume_cache to deprecate and replace with a
single call to disk_driver that contains all of the options. Usage of
the volume_cache option will now result in a message to ui that it has
been replaced, as well as a warning that it is ignored if disk_driveris
set.
The old option volume_cache is only used if disk_driver is not present
(even if :cache is not set - in that case, the hypervisor default is
always used).
Resolves#998
virt-sysprep by default enable the `customize` operation, which will
regenerate the `/etc/machine-id` by the end of the operation. This
cause all newly created VMs by `vagrant up` coming with idential
`/etc/machine-id` and so receive conflicted IP from DHCP.
This PR disable `customize` as per
https://github.com/vagrant-libvirt/vagrant-libvirt/issues/851#issuecomment-520799720
recommendation.
Fixes#851
Allow configuration of various memballoon-related settings. It was
discovered that it may be needed to be able to control the memballoon's
PCI slot/bus location in order to prevent it from conflicting with
other explicit PCI location assignments. For example when configuring
the management network NIC to go to slot 0x05, libvirt would try to put
the memballoon there as well and resulting in a fatal error.
Qemu has supported tpm 2 and the ability to start swtpm. Additionally
it expands the tests for the tpm configuration to ensure that only when
the options cause a change to the domain XML will the domain be updated
on a subsequent start. This change just allows passing through the
necessary config.
Vagrant.configure("2") do |config|
config.vm.provider :libvirt do |libvirt|
libvirt.tpm_model = "tpm-crb"
libvirt.tpm_type = "emulator"
libvirt.tpm_version = "2.0"
end
end
closes#965
Adds support for exposing the correct version via export-subst for
git-archive constructed tarballs to make it easier to consume directly
from source.
Will check in the following order of preference:
- version file
- format string containing "Tag:"
- clone remote to describe commit
When using qemu:///session, it's necessary to ensure the correct
user/group is passed in when creating additional volume storage as
otherwise the default is to attempt to chown/chgrp it to 0:0 which will
fail.
With this in place and recent changes around uri/qemu_use_session,
remove the checks guarding retrieving the storage pool as it is also
possible for it to be created as expected for the session.
Update create domain tests to check for the correct settings such as
storage path and user/group id's passed to the volume create call for
the additional disks.
Fixes: #986
Various checks in the start domain action were accidentally causing a
redefine right after initial create. Update to provide debug output when
the domain needs to be changed to allow capture of the redefines
occurring in the future and to make it easy for the tests to pick up
where the redefine was triggered by setting an expectation on the log
output.
Include a small fix to avoid running strip on what might be a nil object
returned for elements without any text attributes.
Fix a bug where changes to tpm settings made to the config after an
initial domain creation where there was previously no tpm defined, would
be ignored.
Adds a logger double and updates other tests that trigger log calls.
should fix#1176
hashicorp/vagrant#11465 introduced a new option --no-tty which silences
progress bars so they don't spam logs when Vagrant is used
non-interactively. First Vagrant version with this change is v2.2.8.
However, existing code needs to be slightly updated to cooperate with
the new option correctly, otherwise the progress bars spam the logs with
empty lines/warnings instead of remaining silent.
Fixes: #1106
Add default domain start spec test and fix bug triggered by whitespace
mismatch in string comparison that would trigger unexpected domain
undefine and recreate.
Switch to looking up the version from a file with a fall back to get
it directly from git tags if the file isn't available.
The version file is automatically generated by a task of building of the
gem and included in the package release to prevent reading from git.
Should allow the release process to be automated from pushing of a git
tag.
Add support for a pool setting for additional disks, example:
Vagrant.configure("2") do |config|
config.vm.box = "generic/centos8"
config.vm.provider :libvirt do |domain|
domain.storage :file, :size => '20G', :pool=>'default'
domain.snapshot_pool_name='cache'
end
end
this allows to place the virtual machines snapshot in the "cache" pool,
while additional disks are created in the "default" storage pool.
Provide a simple create domain test that uncovered a bug with an
exception in addition to fixing the tests to avoid unnecessary output
when the code sends messages via the UI.
Erubis is dead upstream, therefore it is good idea to avoid its usage.
Also, this allows to drop explicit dependency on Erubis, which is not
specified anywhere and may cause issues when Vagrant changes its
renderer.
Fixes#1142.
Users are allowed to set a LIBVIRT_DEFAULT_URI environment variable that
controls tools (i.e. virsh, virt-install, etc) that communicate with
libvirt. Let's allow for that mechanism to be used here.
Test settings modifying the `@uri` and `@qemu_use_session` variables to make
it easier to update and perform an initial minor refactor to reduce some
of the code currently in use to set `@uri`.
According to the docs, for volumes and backingStores: "The mode defaults to
0600 when not provided. https://libvirt.org/formatstorage.html
Removing this setting here allows the pool configuration to set the default
mode.
- Explicitly disable any SSH multiplexing here so PID tracking of
tunnels works correctly.
- Using `exec ssh` in the spawn forces Ruby to use a subshell (as exec
is a shell builtin) instead of spawning the ssh process directly, which
results in getting the wrong (and dead, as the exec replaces the
subshell) PID to track and clean up later.
- Run the ssh tunnel command on its own process group, essentially
daemonizing it and keeping its PID intact even when `vagrant up` command
is run on shell/consoles without an explicit TTY, such as Emacs Eshell.
Make it easier to tweak some qemu options by allowing passthru of
command line environment variables.
- Also cleans up weird variable indirection used for qemu commandline args
variable through `qargs` in various constructors.
- Addresses some functionality discussed in #776.
Packaging an existing vagrant environment relied on copying the backing
image. Users may not have permission to read this file despite being
able to interact with it through libvirt.
Use ruby-libvirt to download the image instead so any user with libvirt
access can run the package command.
Uncomment NFS prune action
If the user has not configured NFS for any of the synced folders, then
it is likely they are not expecting to be asked for a sudo password on
destroy. Move the test for using NFS to a common module and include in
both the preparing and pruning actions.
Co-authored-by: Darragh Bailey <daragh.bailey@gmail.com>
The fog-libvirt volumes filter behaves differently than the others in
that it returns an empty volume object if nothing is found. This has the
potential to change in the future so guard against an empty result as
well by checking if the assigned first element is nil.
This commit replaces the pattern where the metadata for all volumes,
servers, or pools was downloaded and then searched. Instead a filter
argument is passed to the connection and only the metadata for the named
resource is returned, if it exists. This results in significant speed
increases for libvirt hosts that are offsite or have many volumes.
Warden will call the recover method when env[:interrupted] is set, so
just return instead and let it walk the stack to perform cleanup. This
ensures that during the WaitTillUp action that the user setting to
disable destroy_on_error is respected should the exception occur during
looking for the IP address or SSH connection.
