This role now expects disks to be passed as a list instead of a
comma-separated string.
This commit updates group variables to stop transforming disk lists to
comma-separated strings and adds a check for users overriding LVM
variables. The playbook is also tagged as upgrade-check.
Change-Id: Ia8001c28a8be034ae79a1c584beb40bc66891db3
Bifrost does not perform ironic online data migrations after upgrading. This
can lead to the following error during kayobe seed service upgrade:
The database is not compatible with this release of ironic (10.1.7). Please run
"ironic-dbsync online_data_migrations" using the previous release.
As a workaround, perform the migrations in kayobe, prior to performing the
upgrade.
Change-Id: I8951242ba40158c26e85377254d19085d85cf753
Story: 2004308
Task: 28657
Adds the kayobe-seed-upgrade-centos job, which performs an upgrade of
the seed services from the previous release to the current release.
Change-Id: Ia3eb39cf81cb3618fd94c4456bd576b52098c946
Story: 2004308
Task: 27873
Adds a new command to upgrade containerised services on the seed:
kayobe seed service upgrade
This is largely the same as the deploy command, since there is no
kolla-ansible bifrost-upgrade command yet.
A limitation in Bifrost requires us to remove IPA images before
upgrading the bifrost container.
Change-Id: Ibbe01e972487e69dbb753a67efcf13e3ae0800de
Story: 2004308
Task: 27872
The 'kayobe overcloud inventory discover' command can be used to discover an
overcloud inventory from the nodes in the ironic service running on the seed.
This works well for cases where the nodes have been autodiscovered, and as such
do not exist in the kayobe inventory.
There is another use case where overcloud nodes already exist in the kayobe
inventory. In that case we don't want to run the above command, as it will
create duplicate hosts in the inventory. We do however want the part of that
command that generates bifrost host variables for the overcloud nodes.
This change moves the generation of bifrost host variables to the 'kayobe
overcloud provision' and 'kayobe overcloud hardware inspect' commands,
which are used in each of these cases.
Change-Id: I8a69ccd5192c0c5beeb7efb36c09b2c4643ef150
Story: 2004426
Task: 28079
Changes to the openstack ansible modules and client SDKs have rendered our
authentication for ironic invalid. Bifrost now provides a working clouds.yaml,
so use this instead of the fudged endpoint/token setup in env-vars.
Also updates some use of the ironic client to the openstack client, and adds
support for the new 'inspect wait' ironic state to avoid surprises during
inspection.
Change-Id: I15ea388b6df8ced9cc0e0eceed8dec8aa8f57a9b
Story: 2001864
Task: 28078
Kayobe writes out several host variables to the Kolla ansible inventory
files, etc/kolla/inventory/seed and etc/kolla/inventory/overcloud. These
include ansible_host, and network interfaces such as api_interface,
ironic_dnsmasq_interface, etc.
In Ansible, these should have a higher precedence than the kolla ansible
group variables in ansible/group_vars/all.yml that set the defaults.
However, in Ansible 2.4+, if the host has the same name as a group that
it is in, the group variables now take precedence, meaning that it is
not possible to override them.
This was observed when using the kayobe-config-dev repo for testing,
where the seed host is in the seed group.
Admittedly ansible does tell you not to do this: [WARNING]:
Found both group and host with same name: localhost
The solution used here is to use a separate host_vars file for each
host. Alternatively we could enforce that hostnames and groups do not
overlap.
Change-Id: I349c8279d85a591689ac8108bce14d96889440fe
Story: 2004418
Task: 28065
The dell_switch_delegate_to parameter was removed from the dell-switch
role in commit da75d6b734fa0451738993d4d6c9beab9b4640b2.
Change-Id: Ia70d48d417d68140ea9e3abd1409c6b0945e638c
The dellos Ansible modules do not accept a `transport` key in the
provider parameter anymore, even in Ansible 2.4.0.
Change-Id: I4bdaa337cb9c6cb93f36338a191a9aa78e2c13f0
Story: 2004588
Task: 28464
If a public FQDN is defined, use it instead of the virtual IP address.
When using TLS, certificates are generally tied to a hostname rather
than to an IP address.
Change-Id: I0afef32a0f4030bba613febdf266e2ccd863512d
Ansible Galaxy switched to semantic versioning. In order to get it
to import the latest release, a tag following this scheme was added
after a minor update to the grafana-conf role. This appears to have
broken the accessibility of the original version. This commit moves
to the later release to work around the issue.
Change-Id: I5a0ba9e18bc5db54f899b545b88c319c7922bf9e
The administration guide has become a bit bloated, and not well
organised. This change creates new subsections within it.
Change-Id: I8aa133c183fb6b7d87a86d5ab66cc7faa5e64112
Story: 2004337
Task: 28082
The Juniper device support allows ansible_host variable to not be set
for a switch, in which case the inventory_hostname will be used instead.
This change applies this default consistently across all switch types.
Change-Id: Ibc1e673a6b988f2e8d7bb6c62d30ccdb7ce96351
Story: 2004358
Task: 27958
previous_release variable is used for two different purposes in Zuul config.
Rename one to reflect its flagginess.
Change-Id: Ib7833b29ab2959db7c7ff85b90fb8a4f5082d864
This small update is to show users where to put their manila
configuration file for kolla-ansible.
Change-Id: I050bb6e388c8b82de05e4deeaa357b98525c004e
Story: 2004414
Task: 28058
This commit enables the merge of additional manila.conf
as seen with the other projects.
Change-Id: Ia2209a9c82215e49f1ef5a79ecafbebf2b204627
Closes-Bug: 28058
Kayobe should be able to handle extra configuration options in Octavia's
main configuration file, in a similar fashion to other services. This
commit adds support for just that.
TrivialFix
Change-Id: Ibd91a90a5278bcd8963fb06874a74f71b1882d89
Previously we checked for false in the result. The behavior
of Ansible has probably changed, as failed is always in the
result dictionary - just with the value "false". Tested
with Ansible 2.5.11.
Change-Id: Idfa2f33dd6641031ea8a287d74b6dbbfa578c9ab
We are seeing a few overcloud deploy/upgrade jobs timing out. The jobs
appear to be progressing, so killing them isn't particularly helping.
Minimal random sampling suggests this affects the OVH cloud.
Change-Id: Ibce5457edf7a5f06e834aa587b5d4b5b6c69b8b6
