This, alongside the "df -h" output, may provide some useful information
about the /opt content (not as precise as a `du -shxc /opt', but at
least it's already something). It can then be compared with the
existing `df -h' information located in the zuul-info/ directory.
The goal here is to understand the actual impact of this task.
Change-Id: I201eba6ec6db9f0caaaabb0a3d15b0894125c7d6
Disables translation sync with Zanata on Zuul part
since OpenStack-Helm would like to cancel translation.
Depends-On: Ica81dfefaf848f03c972db79786735ffcff99351
Change-Id: I95db692d0243af0b49a8c04c2fd18f01914dd743
tools/generatepot-rst.sh has an awk filter to drop UUID from POT files,
but it is unnecessary now. It was introduced to strip UUID information
which was added by default in Sphinx <1.3 [1] in openstack-manuals.
Sphinx >=1.3 does not output UUID information by default [2],
so is is no longer needed.
In addition, (g)awk 5.0 or later (adopted in Ubuntu 20.04 or later)
complains the current awk regexp. I think it is the time to drop it
rather than fixing the regexp.
[1] 993647f316
[2] https://www.sphinx-doc.org/en/master/usage/configuration.html#confval-gettext_uuid
Related-Bug: #2035226
Change-Id: I57ce0c7d662a4cfd84d9647645a0c6ed9c56ac91
OpenAFS 1.8.8.1 doesn't build on centos 9-stream, so this is the next
available release. Even though it's a pre-version, it's got to be
better than cherry-picking patches.
Change-Id: I02ef7fd5500cd43b3fdc210e88d133f50a93c017
Pin ansible-lint<6.5 for now due to the
https://github.com/ansible/ansible-lint/issues/2320 regression which
ignores our skip tags.
Also add the zuul.d directory to the list of directories to avoid
linting, since those YAML files aren't Ansible files and it doesn't
seem to know not to try parsing them anyway.
While we're here, "fix" a task name not starting with an upper-case
letter, which for some reason ansible-lint now seems to think is
important. If we decide to continue using ansible-lint, it's
probably going to persist in complaining about that.
Change-Id: I46826210dfa4442af5c36d1439d1a1047ca0ba90
Lots of builds fail with
$ grep -r "invalid command 'bdist_wheel'" * | wc -l
388
I think perhaps we need wheel in the venv, as packages don't specify
the depdency explictly?
Change-Id: I2ed1b07e9e6dd5513f30b3007ee187da7665bf35
Now that we have shown this working with a test PPAs, this pushes the
packages to the repos used by production.
Change-Id: I170d38c311a37880081c8b195ce3a99075907f29
Use the generic role from I11e2bdcbf2894d86105d75c3d3580aefb6aa18fc
and make a generic publishing role too.
Remove the openafs specific bits that are no longer used. Move the
description of the OpenAFS jobs above the job definition.
Change-Id: Ie61b82903644605d07f510ad1071fa551e7086c9
This is a reworking of the openafs package build script to be more
generic. This will be used by vhd-util in follow-on changes.
Change-Id: I11e2bdcbf2894d86105d75c3d3580aefb6aa18fc
Per the notes in Ie00c42857e7a541aa1ccb8fe480756c1e7a38dc4 having one
debian/ directory checked in here doesn't work across multiple
distributions. Rather than trying to keep mulitple versions around,
or munge files, we've split the debian/* and source generation into a
separate repo.
This makes the job more generic to build the debs from this repo. See
Ie66d3b1e39ef9fa714b1dabdb7eb61cc43538587 for an example of using this.
Change-Id: Ia250acf64b76ef4e1b981f2ebb1febe6e063157c
As noted inline, if the source already exists we need to use it from
the primary archive. Do this for 1.8.8.1 but leave notes behind on
building non-packaged versions.
Change-Id: I6d637c81dd4a8d712b5fa84722d7559178e40de2
This imports the latest debian/* from the upstream openafs packages,
and builds source packages meant for upload to the PPA we use for
openafs.
Over the years we have found that maintaining our own PPA for this is
useful. We have hit bugs that require immediate patches to be applied
and rolled out to production, and having a private build facilitates
this. Generally we hope to have no delta to the upstream packaging,
which is currently the case.
Currently we have a fairly manual process to produce and upload these
debs (once there, the source is obviously available). This allows us
to keep a little clearer view of what we've changed or are working on.
This doesn't publish anything, yet. We'll work on that after this;
the idea is that the publishing job will consume the artifacts this
returns.
Change-Id: I07780595331da29af9f72185cbb3499210e89c19
As noted inline, platforms are starting to decide to deprecate
virtualenv for "python3 -m venv". Put in logic to use venv in our
Python 3 paths.
Change-Id: I46778642647d7be995d476fb567a9943cae66b8f
Make sure the pip in the venv is upgraded past the system-version so
it can install all the latest requirements.
Change-Id: I601aca1507d16912e08ee9aba8842bc205ea2a91
Older ansible-lint has an uncapped dep on 'rich' and newer rich breaks
ansible-lint. New ansible-lint has addressed this. We update
ansible-lint to get those fixes.
This adds task names to a number of tasks to correct a linting error.
Change-Id: I38284fef54213e0fceb1b348d5839129cc15148e
The swift team want to translate docs into other language.
For docs translation in zanata, I add swift to DOC_TARGETS
Change-Id: I0d9f46d576edf10117d73526505d1a59b40886dc
neutron-lbaas is retired now and we can
remove it from zuul jobs. It also retire django_openstack_auth
in this patch to unblock the gate.
`django_openstack_auth`` code was merged into the main horizon
repository during the queens release. This repo is already
deprecated and now we can retire it. This patch removes all
references of ``django_openstack_auth`` as mentioned in [1].
[1] https://docs.openstack.org/project-team-guide/repository.html#retiring-a-repository
Change-Id: I5310983bed6c5b1618b7e131996e09571be9f13d
This change fixes the pdf-docs job by installing
tex-gyre to provide the missing tgtermes.sty style.
Change-Id: I885ed9153d9ccd9d3a486b5fc573238520474592
Likely due to many less Python 3.9 wheels the Bullseye builds take too
long. Given this is a new distro, limiting the builds to the latest
releases is a practical solution here.
Add the bullseye jobs so they test in the gate.
Depends-On: https://review.opendev.org/c/openstack/project-config/+/798029
Change-Id: I9130014837d297fad6802c99da417dbb8f3a8b76
Horizon is switching Chinese locales from zh-cn/zh-tw to zh-hans/zh-hant
in horizon and horizon plugins.
This change follows Django change which happend more than five years ago
and the new language code form is commonly used.
Language codes in Zanata are not changed, so we need to covert the
relevant language codes in propose_translation_update.sh.
It allows us to avoid doing it manually and also allows us to
switch these language codes almost once across horizon and plugins.
The detail is explained in the mailing list post.
http://lists.openstack.org/pipermail/openstack-discuss/2021-February/020169.html
Change-Id: I1420508f0513c52866aa637784a51f7c22f6d204
The openafs 1.8.7 release fixes a fatal bug in 1.8.6 that occurs after
January 14, 2021. Bump up the version to 1.8.7 to package up the fix.
Change-Id: I83fea73f0f68d297ed95b14c29cc1e6bb12d6d5d
Per the issue linked inline, the new pip resolver downloads (and
reports) wheels that it ultimately seems to discard itself. As of
this writing, it's unclear exactly what is going on. We can simply
ignore missing files here with -f to safely work around this.
Change-Id: I30700e0a0a242f9d7d2563c989a1d25eb6acd6e4
In [0] the default swap size was reduced from 8G to 1G, update the
readme to match the current default.
[0] I5ce9064a5546d85ab6f41c4e60aa56fa56ff47c1
Change-Id: I5fa3ea0b22cb9e4ce3aa140409f3ee23d3cade35
PDF builds need to continue working on Ubuntu Focal. Installing
texlive-full is an easy answer but pulls in 3GB worth of additional
packages and is causing jobs to frequently hit build timeouts.
Re-add just the focused dependencies needed to keep this working
without requiring the kitchen sink.
This is a partial revert of 3f5d26a1c0
which initially added texlive-full in an attempt to solve PDF build
issues.
Change-Id: Id8f50ccd75bbfe5ac126c61fe8e9223e0b6ae2b7
A recent kernel update in Fedora has highlighted that fallocate produced
swap files on ext4 partions are no longer accepted by swapon:
swapon failed if the swap file is created by fallocate in ext4 partition
https://bugzilla.redhat.com/show_bug.cgi?id=1827115
This change assumes that all distros will eventually encounter this and
removes the fallocate task from the role and defaults to using dd. As
this is considerably slower the overall size of the swap file is reduced
to 1024MB.
Change-Id: I5ce9064a5546d85ab6f41c4e60aa56fa56ff47c1
tox lives in {{ tox_executable }} which is set by the tox role,
so don't use the hardcoded 'tox' name, it might not exist.
Change-Id: I4c93fb838d6476357959a2d4459dede80caa2ad2
Because there are many more wheels to build on arm64, we need to
reduce the arm64 builds to just latest two branches to have a
reasonable build time.
Output the script output to logs; if the job gets timeout killed we
still get some log output.
Remove --progress from parallel as it just floods the logs with output
that isn't useful to store.
Remove --verbose from pip to try and reduce logs. We can still see
failures but currently they are like 500mb compressed.
Zip and remove the logs in post so we don't copy over uncompressed
logs if we are timeout killed.
Change-Id: I5fba52af7fcafb12d0d2c4b416d5e0b7e4feb238
It turns out there are many issues doing a recursive download from our
log servers. Tar the results so we can pass it through a single file,
and extract that file in the publish jobs.
Change-Id: I2b54c061ae3563a8244a39d08e2fe49c2f885d2e