When generating index.html files for uploading to Google Cloud
Storage, link directories to their index.html files because
(except in certain circumstances) Google does not automatically
serve index.html files at directory URLs.
Change-Id: Ie854de328057ed080c80df9328163da98014ea36
This allows a deployer to indicate that they prefer the Zuul
dashboard to link to "dir/index.html" rather than just "/dir" when
displaying directory entries in the log browser. This helps with
log storage systems which do not automatically serve index.html
files at directory urls.
Change-Id: I0444798c689667e51d8761a20b661f8a9b90323d
High Availability packages have been added as part of core CentOS
repositories in 8.1.
This patch add this repo and enables it as it provides corosync and
pacemaker based HA solution and some other packages required for it.
Change-Id: Idbddd81f251c1ade97892128e52f9214420bead7
We have to be careful about avoiding outer loop loop_var conflicts in
ansible. Because the zuul-jobs roles are meant to be reconsumed
elsewhere we should not use 'item' loopvars and instead set them to
something a bit more unique.
We use a zj_ prefix to try and be unique to this repo and document this
convention.
Change-Id: I20b9327a914890e9eafcb2b36dc8c23fb472bc8f
If we install our own local version of tox, we should set a fact
with the tox_executable so other tasks can leverage it.
Change-Id: If6895bbb898261e88c0e3083d21210209f79995f
The upload-afs role is actually fairly specialised for uploading docs
with its root-marker scheme. As we're moving more parts of our
infrastructure to publish via AFS I think it's worth making this quite
explicit by renaming the upload-afs role to upload-afs-roots. This
should be transparent as we call the renamed role, and once callers
have been migrated we can use the usual deprecation process.
Add a partner role, upload-afs-synchronize, which is more of a
straight copy with options to synchronize. This will be suitable for
uploading tarball and release artifacts. This is similar in concept
to the promotion job used by zuul [1], but in this case will be
focused on the extant publishing jobs in project-config which don't
currently use the artifact-reporting/promote process.
[1] https://opendev.org/opendev/base-jobs/src/branch/master/playbooks/artifacts/promote.yaml#L39
Change-Id: I493d8829e3fd98e84f1b8f0e776e7ba41abf16c2
Story: #2006598
Task: #38600
The default loop variable 'item' cannot be nested. Because roles like
collect-container-logs may be called from higher level loops that use
'item' we override the loop control var to a name that should be unique.
This should probably be done to all of the other roles in zuul-jobs too.
Change-Id: I2f647596dd40c662aa7447e3d3c8844a77c6b109
Explain the intention of this role and the details of .root-marker,
some of it is copied verbatim from the infra-specs repository design
document.
Change-Id: I6d9124c7eafadc4b53626ff2fef6be6399f61947
Don't try to create change tag if zuul.change is not defined that is the
case for periodic builds.
Change-Id: I5a7e02caf0f502e90e0477daeee868bf06dfbb5c
Use a proper link to the referenced role and mention in the short
description that this is related to OpenShift.
While it might work with any Kubernetes, it uses an
openshift tool, oc.
Change-Id: Ia4fba00c8a3adec4dc26a2d32a21df8ff548a120
ansible-2.9 does not allow "|skipped" anymore, it needs "is skipped"
instead, fix the only usage.
See also Ic0c93f2ea896ead76922694e6150882d8a280daf and
Id6c62da8f8a94c270d8a5d5981683a9d8c4e7ad6.
Change-Id: I597c92d2f6dce31dab1ac4bf52bb534c8146b33b
This seems to introduce the following error with Ansible 2.9
"The conditional check 'not ara_generated | skipped' failed.
The error was: template error while templating string: no filter named 'skipped'.
This reverts commit 262c16b30583f634588f5a52bd7aab944687a8a4.
Change-Id: Id6c62da8f8a94c270d8a5d5981683a9d8c4e7ad6
The
command -v pip pip3 | head -n1
introduced with Ie50928c9b782ea84db916bb1441567e1206ff466 has a very
subtle race; if "pip" and "pip3" exists and there are two lines
output, the "head -n1" will exit and depending on scheduling the
"command" might write to a broken pipe (this manifests as exit code
141).
Move this to a more explicit if statement.
Co-Authored-By: Jens Harbott <j.harbott@x-ion.de>
Change-Id: I80823a7bc6351925d6f0b20bdebca3eafef0b27d
The sphinx .doctrees directory is not needed for publishing. By default
sphinx puts it below html. Some projects add to sphinx-build a "-d"
option to put the content elsewhere and not publish it.
This change assures that a .doctrees directory below "html" will not put
stored as artifact and published later.
Statistics: In all of OpenDev, we have:
* 511 repos putting .doctrees elsewhere and thus not publishing it
* 631 repos no caring and thus publishing .doctrees
The size of .doctrees directory is not neglectable: 2MB for
system-config, 14 MB for nova.
Change-Id: I63b581164aab84352d9e80278cf4a8d634ff28af
When returning the subunit test result as an artifact it will be
directly linked on the build result page.
Change-Id: I1a83288ac0857333f87cd085b3e0243bd3b92fc2
Fixed issues failing to install tox on python3 only systems which have
only pip3 executable and not the pip one.
Needed-By: https://review.rdoproject.org/r/#/c/24584/
Change-Id: Ie50928c9b782ea84db916bb1441567e1206ff466
This change enables using fetch-tox role along with
the fetch-output role. By default the role still synchronizes
artifacts back to the executor.
Change-Id: Iffe5ffc52ee4c765b2e23ab8b3235b6bc7c510d8
This role runs Helm template of a specific chart and then makes
sure that all resources become ready afterwards.
Change-Id: I62e160eac673d13bd1018b58fd36cae54e1482b4
We use this in a few different places and it's really useful
to collect all the logs of all containers.
Change-Id: Idc46a47f444bf48cd040f4f9724f3a6ee8bc8f8e