Previous version of nfs-provisioner was 2 years old,
uplift the version to possibly fix current instabilities,
such as data-lose for MariaDB and PostgreSQL pods.
This adds new variables to the 005-clone-dependencies script in
Airskiff to be able to turn off cloning for the Airship repositories.
The intended use case of this is so the Airskiff gate can avoid cloning
the repository it is currently testing.
Add explicit glance and heat overrides, this fixes problems
when two manifests with identical names are present,
one manifest comes from the type level, another one comes
from the site level, this leads to random manifest being
picked up during service deployment by Armada.
The UID:GID under which commands from within containers are run does not
necessarily match UID:GID set in container. This patch enforces run of
containers with process UID:GID matching current user's UID:GID.
Creation of directories is documented, so that user's wouldn't get
Configure Airship Seaworthy site specific parameters
when running tests. This would allow to make required
changes to default values in ./tools/openstack script
which are currently Seaworthy specific.
Note: all scripts in tools/deployment/aiab directory have been
moved into this repository as is (except a few changes to make
them work), they require a heavy refactoring that will be done
in separate patch-sets.
* Add a virtual single node manifests that are based on a sloop type.
* Use NFS provisioner instead of Ceph.
* Update tools/openstack to be non-seaworthy specific, use a default
region name and auth url.
* Make type/sloop/config/common-software-config.yaml to be site specific,
to allow to configure custom region_name.
* Remove max-pods-per-cpu parameter for kubelet, treasuremap needs
to support a diverse set of environments, without constraints
on the number of available cores. Max pods configuration parameter
is still present and helps to mitigate problems when kubernetes
unexpectedly starts a large number of pods.
Currently the manifest targeted by armada in the
030-armada-bootstrap.sh script is hard-coded to "ucp-bootstrap". This
patch set updates the 030 script to make drive the target manifest
off of a new environment variable "TARGET_MANIFEST". By default
"ucp-bootstrap" is still used, and existing functionality should not
The motivation behind this, is it allows us to use the armada
bootstrap script to do more than just bootstrap the ucp components,
which may prove useful in zuul gate jobs that use treasuremap.
Since support for additional operating systems has been added to some
components, the image tag "latest" is no longer used by their Makefiles.
This commit changes the tags for the aforementioned components to use
the new default tag and updates the job name to reflect the use of
One of the first things the tools/airship utility script does is run
dpkg and apt. Because these commands were not being ran with sudo,
the airship script could only be ran as the root user. Adding sudo
so any user can run the tools/airship script.
Added service specific version of helm toolkit for all OSH services
in treasuremap global chart and version definitions. Kept the current
version of htk for all services to limit the changes in this commit.
Globals have logging override defined which is now different from
openstack services values in openstack-helm.
Openstack-helm addressed issue of missing logs related to warnings and
above for root logger which is catch-all logger when component specific
logger is not explictly defined.
OSH review: https://review.opendev.org/#/c/639711
Dropping logs in case of error or warnings can be problem for operators
in triaging an issue. We should not be hiding warning/error/critical logs
even when specific component logger is not defined as there can be cases
where each component is not defined in logging definition. See example
from OSH review.
In addition, is there a specific reason to have logging override for OS services
in airship side as they appear to be quite similar to defaults defined in
openstack-helm side. So may be we should revisit its need and address in
different patch if needed.
Currently, many of the charts in the sloop type are missing labels that
are required to enable layering at the site level. This change adds the