Signed-off-by: Samuel Verschelde <stormi-xcp@ylix.fr>

In 0a5ea64 the method was moved from
BaseVM to Pool, but an intermediate version had it moved to Host, and
the final version missed the conversion on one call site, breaking
tune_firstboot tests.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This in INFO level makes user think the test is blocked on
xapi-wait-init-complete while the problem is with reaching pool members:

 Sep 06 12:50:28.791 INFO Wait for XAPI init to be complete on 172.16.210.12
 Sep 06 12:50:28.792 DEBUG [172.16.210.12] xapi-wait-init-complete 60
 Sep 06 12:50:29.221 DEBUG [172.16.210.12] xe host-list --minimal
 ...
 Sep 06 12:50:30.503 DEBUG [172.16.210.12] xe host-param-get uuid=0aea61f4-c9d1-4060-94e8-4eb2024d082c param-name=address
 Sep 06 12:50:30.926 DEBUG > 172.16.210.11
 Sep 06 12:50:30.929 DEBUG [172.16.210.11] cat /etc/xensource-inventory
 Sep 06 12:52:47.216 DEBUG > ssh: connect to host 172.16.210.11 port 22: Connection timed out

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Samuel Verschelde <stormi-xcp@ylix.fr>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Getting a vm_booted_with_installer starts with inserting a CD, will be
progressively enhanced.

ISOs are specified using a filename, or a download URL used through a
local cache.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This is a first step for answerfile handling, to be able to run a
first real installation.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Currently relies on the installer fetching the answerfile from the PXE
server, to ensure the latter's ARP tables are populated with the
obtained DHCP address.

Monitoring the install progress is done by repeatedly polling with grep,
which is far from ideal performance-wise, but avoids adding too much
complexity.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Detection of host IP till now relies on the fact we download the
answerfile from PXE server.  Once we take this file from the ISO this
network traffic won't happen so we need some other mechanism to fill the
server's ARP tables.

test-pingpxe.service is installed in install.img by iso-remaster.

Since it is difficult to wait until the IP has been assigned before
launching the service, make it ping continuously until we can reach
the PXE server.

One the installed host we will set a static IP instead.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This is necessary to get rid of old ARP cache entries that would match
our IP to the MAC used by the VM clone in a previous test.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This process has several steps:
- building of a data structure holding all of the answerfile data,
  from a customizable base in data.py and from tests-specific items
- serialization as XML to be read by host-installer
- necessary changes to the ISO for host-installer to use it

We now have to explicitly enable the network during access (was implied by
the use of a remote answerfile).  Similarly we now rely on the test-pingpxe
service, as nothing else would otherwise populate the server's ARP table.

This is needed so:
- different tests can use different parameters without the need for
  provisionning every answerfile to be used
- tests can dynamically add contents for their own needs, before the
  XML gets actualy written

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This is not ideal, as we would need to encapsulate machine boot in a
fixture - that will be fixed when we switch away from
pytest-dependency.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This protects against using results from incompatible test by mistake.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

The key used to locate a VM image in the VM cache depends on the test repo
commit hash, to protect against using results from incompatible test by
mistake.  But the commit hash can change for many reasons that do not
influence the parent tests, so this provides a way to use known-equivalent
test outputs.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This will avoid duplication of logic for selecting parent test.

Since dependencies are now generated by fixtures, they are not taken
into account by pytest-order any more.  Looks like some heavy surgery
would be necessary to get that back.

Especially useful for images out of CI, which today are not signed yet.

Changing test parameters may require changes to hardcoded test sequences.

Adding only a single test sequence here to server as example for quick
testing, as further commits will make many changes to test names.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

FIXME: those xcpng versions want the ed25519 key, while the alpine
templace still only has the rsa one.  Single .ssh/config seems hard to
configure given use of DHCP.

The installer generates UUID for host and its dom0 in xensource-inventory,
so the result of a costly install test could not be reused for multiple
hosts in a same pool.

This service will be run once and before xapi ever starts, to override
those UUIDs with brand new random ones during firstboot.

FIXME: should be moved to test_tune_firstboot instead

This is just base infra, different machines have not differences yet.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Uses a helper VM to modify firstboot data in installed disk

test-sequences update:

 git restore -s HEAD -WS --ignore-unmerged tests/install/test-sequences
 git clean tests/install/test-sequences -fdx
 sed -i '/test_firstboot_install/ i tests/install/test.py::TestNested::test_tune_firstboot[None-' tests/install/test-sequences/*lst
 # plus manual copypasta for rest of line in:
 git st --porcelain |grep ^.M| cut -c4- | xargs $EDITOR

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

When the ISO is copied to the SR it happens that it is not available
right now. So rescan the SR to ensure that it is available.

Signed-off-by: Guillaume <guillaume.thouvenin@vates.tech>

FIXME: rescan after copying rather than on each use

clone+start:// will be used to implement--hosts=cache://...

clone:// itself is not yet used directly, but as the "base" protocol upon
which clone+start build, it seems logical (and basically free) to implement.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

FIXME:
- installer_iso fixture could just use iso_version param
  ... except for gen_unique_uuid
  => turn into optional iso_remaster marker?

Add a new parameter to allow booting using netinstall. If an ISO only
supports netinstall you can specify it in the data.py by setting the
option `net-only` to True. This option is set to False by default.

Signed-off-by: Guillaume <guillaume.thouvenin@vates.tech>

FIXME: net-url should not be required for every ISO

FIXME: config.py seems a bad idea after all

…llations

…allations

FIXME: duplicated logic should be factored out

This is not meant to be a fixture anyway.

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

This suppresses hostkey checking, making the script more suitable
for non-interactive use, and makes it easier to understand when something
goes wrong.

FIXME locate "lib" more properly

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Tests for generation of the reference files, and for checking against them.

byte-compiled version of /etc/xapi.d/plugins/vmss

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

Signed-off-by: Samuel Verschelde <stormi-xcp@ylix.fr>

When using --hosts=cache://... it is much too early for such a check, so
we have to give the info manually, and the script would not allow this.

And honor it not just for "collect".

…llations

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>

FIXME: dependency on tests from another module don't seem to get properly
located by pytest-dependency (--ignore-unknown-dependency required)

FIXME: how useful?

FIXME: had to keep a copy of perform_upgrade() after its functionality
was moved to vm_booted_with_installer fixture, must find a proper solution
for that

Useful to get the debug output from commands.ssh()

Signed-off-by: Yann Dirson <yann.dirson@vates.tech>