The previous uwsgi invocation assumed that the uwsgi binary is in the
same directory as the project binaries are installed into (probably
/usr/bin). That may not be correct -- for example if using a packaged
uwsgi on Fedora the binary will live in /usr/sbin/uwsgi (not /usr/bin
where the project files from pip are).
Switch the invocation to just find it in the path.
Additionally, grenade is disabled, otherwise we cannot land this change
without fixing stable branches first.
Change-Id: I7f627fa0bb4850f4839b75df611fa5ab241f1219
Sometimes sushy-tools hits worker timeout in gunicorn when handing
a virtual media ISO, causing a request to be aborted and deployment
to fail. Increase the timeout from 30 to 90 seconds and add some
concurrency to ensure that more than one deployment can run.
Change-Id: I9e951aaaa13981602151bff3f2eebfa338113f6c
Runs the BaremetalSingleTenant tempest test, which provisions
two nodes, with slaac addressing, all provisioning traffic e.g.
iPXE, ironic api traffic and iscsi happens over IPv6 on the
provisioning network. Nodes are then deployed onto a IPv6
tenant network, where tempest verifies connections between them
with ssh (again over IPv6).
Co-authored-by: Derek Higgins <derekh@redhat.com>
Depends-On: https://review.opendev.org/722663
Change-Id: Id928df1cbd1131427ab6a911856f48f18e67f7cf
Now that I6ccd646f1c22a99bed0bebf6e363d2190241b667 is merged, devstack
defaults to 0.5.1, and we no longer need our special logic.
Change-Id: I13d3a4ab7dc330b22d886d85eae94b296e6cac4a
Since IRONIC_RAMDISK_TYPE is default to dib, we should have a
default value for IRONIC_VM_SPECS_RAM that is compatible.
Increasing to 3072 since is the default value we are using in our jobs.
Change-Id: I7b3dab51fa332bbd27a439fd38ebec74d94eb839
Fixing configure-vm script and vm.xml with this patch because for
latest OS like centos8 - older python version and older qemu machine
type is causing issue.
* For configure-vm.py - Replacing python with python3 in environment
path as python3 is the default in many latest OS.
* For vm.xml - Replacing Machine type from pc-1.0 to pc.
- pc-1.0 is old and unsupported machine type for Centos8
- Moving machine type from pc-1.0 to pc will help as pc have alias to
supported qemu machine type
Related-Bug: #1875681
Change-Id: Icd7e98ceb8859f730a32981a150f2730bde9836b
It's supposed to include the patch we need, and using the final
release allows us to benefit from caching of images by the infra.
Change-Id: Iabef3fa7e3af7fb6fd19973d6a44908d5a3ce93b
The default value of "netboot" was introduced to this configuration
variable as part of commit 93f947c852409af2c56a499428b09ff69ab345a1
in Ocata release.
This patch changes the default value of configuration parameter
'[deploy]/default_boot_option' and devstack variable
'IRONIC_DEFAULT_BOOT_OPTION' to 'local'.
Change-Id: I9bf56a7088281bbe20b8b6c2e47c6ab6559bfea4
Story: #1619339
Task: #10505
On non-Ubuntu systems, when gunicorn is pip-installed on Py3,
command name remains `gunicorn` rather than `gunicorn3`.
Change-Id: I5b8cb974d900bd3bdf7ec08adb3e6586494a26a2
We limit the use of tinyipa to the most intensive jobs:
standalone
grenade
multinode
Also leaving ironic-tempest-ipa-wholedisk-bios-agent_ipmitool-tinyipa
running for now as there could be an issue with the
ironic-python-agent
Story: 2007356
Task: 38908
Change-Id: Ieb1ea1dbeeb68b15e9281d154358336006b62214
The agent needs to be able to understand
if the agent token is supported and is a
mandatory feature of the ironic deployment
as that can alter some of the behavior of
the agent itself.
Also adds documentation on the subject for
administrators.
Story: 2007025
Task: 37821
Change-Id: Ic635e0ed6d378d6a34a4a82e66ca647eee33bc26
Centos 7 pip seems to install the rootwrap filters to /usr/local
despite its data path resolving to /usr.
Change-Id: If2496e9c143400dcce017a79cf179768951eb137
The file is included in the ironic-lib module and there is no
difference (beside some comments) in the files.
For packagers, this solves the problem that ironic-lib and ironic are
currently not installable together given that both packages contain
the same file.
Story: 2006604
Task: 36772
Depends-On: https://review.opendev.org/684771
Change-Id: I2d800d4c2616ccd9d9f12a2fe59e259def15dc6b
Remove the dynamically registered ipxe_enabled option and say goodbye.
Further extracts common bits to the PXEBaseMixin, tuning tests here and
there.
Story: 2007003
Task: 37779
Change-Id: I7c1b2a984d45bd63b4e95b62ce02960924c2ce17
It seems, however variables are getting set with grenade
in native zuulv3 jobs, the environment variables are no
longer comparing as they once did in standard bash jobs.
What worked previously, was [ "$VAR" == 'value' ], however
when comparing the newer grenade node jobs we see things like
[[ subnode != \s\u\b\n\o\d\e ]] returning true when single
quotes are used.
Comparison in the logs to where double quotes were used for
functionally the same comparision, revealed proper processing
and execution as expected.
As such, changing our single quotes to double quotes where
applicable.
For the record, I also quickly checked codesearch and it looks
like we were the only project to use single quotes for this
comparison.
Change-Id: I2db2e870e2c5f32aa061af025ee7ce12c4f7c049
Building images may not work without it, and it's only automatically
installed in the CI if diskimage-builder is in required-projects.
Change-Id: Ic0198bc02b1aa9a960aa08c224598d7e1e809012
This switches the piece of code turns on multicast [1] to run with
python3.
[1] https://review.opendev.org/#/c/665221
Change-Id: I12d3764d087fa19aca5b285cbd63a19b1770bb41
We are about to drop pxe + ipxe_enabled=True combination for iPXE
testing, so let's make it explicit in the jobs.
Change-Id: I0e99e6336aa47857c06ba2f6869e66580159a9be
Story: 2007003
Task: 37779
This patch fixes the issue if IRONIC_BUILD_DEPLOY_RAMDISK
is set to false and IRONIC_DEPLOY_ISO is passed then it
will not throw the error.
Change-Id: Ic95f6e75d51db5cb60098bb131698cf44e772366
Story: 2007002
Task: 37777
Adds a job to boot a node over Redfish virtual media in legacy
BIOS boot mode. For sake of unification, similar PXE job has been
renamed.
Apparently, libvirt-based Redfish BMC emulation requires node
power-cycle for libvirt domain configuration changes to take
effect.
Change-Id: I2c7fc26b082db6eeae1abcd67362884a9c72c109
When configuring VMs in devstack booting in UEFI mode, make sure
that each libvirt domain has its own UEFI NVRAM store. Otherwise
funny things can happen ultimately leading to EFI shell popping
up crying for help.
Change-Id: I03eddde04211ceb6f0bd9438fafab34c515c4fa9
the syslinux tools are in the syslinux package on
rhel7 and friends, but in rhel8 have moved to
syslinux-nonlinux
Change-Id: Iea74c2be6fbd738f35f178fa189384dd37579277
Apparently, now days system boot chain is tempted to display
fancy things on the console. While enjoyable, in non-text mode
libvirt guest console can't make it to the log file.
This change tries to ensure pure text mode kernel console in
PXE and Redfish booting by disabling any graphics in the VM
and in guest kernel.
Note that it's not only kernel that can control display mode,
but EFI boot loader as well.
For getting on VM's text console in devstack one could use
`virsh console --devname serial1 node-0` or just add `<graphics>`
section to `virsh edit node-0` XML.
Change-Id: I14f78bcef4173eeaf659ee589ca938e2698e3505
Correcting the variable name for virtual bmc instances during
devstack unstack ironic cleanup phase.
Change-Id: Ie35a45e0806f9da07d121900831c6a61e33d23e5
When devstack boots a node over virtual media in BIOS boot mode,
`isolinux.bin` boot loader needs to be present for grilling hybrid
ISO images.
This patch makes devstack pulling necessary packages and configure
ironic accordingly.
Change-Id: I2f4e582aa05fc54610a4c43695124219714368c2
This feature is time-consuming and prone to race conditions. This
changes switches it off by default, leaving it on in two jobs:
the one running on Nova and the standalone one.
Change-Id: I7c50cb12065da05dad4ece0918883748ee4617e0