Skip to content

Deploy the next lane with Tentacle - #2961

Merged
berendt merged 2 commits into
mainfrom
next-lane-cephadm-tentacle
Oct 2, 2026
Merged

berendt merged 2 commits into
mainfrom
next-lane-cephadm-tentacle

Conversation

@ideaship

@ideaship ideaship commented Sep 30, 2026 •

Copy link
Copy Markdown
Contributor

testbed-deploy-next-in-a-nutshell-ubuntu-24.04 is the lane for the next OpenStack release, 2026.1, and it pinned Ceph to reef. reef is not the Ceph this lane should be validating against 2026.1.

This points it at Tentacle, which osism/release offers as a flavour for latest. The tempest sibling inherits from this job and follows.

The Ceph backend is not set here. python-osism selects it from the OSISM release: the latest track, which this lane runs, deploys with cephadm, and ceph-ansible has no Tentacle path in any case. Naming a backend in the job would only restate that, and the nutshell collection would ignore it.

That selection is not confined to this lane — every nutshell lane on the latest track deploys with cephadm, including testbed-deploy-in-a-nutshell-ubuntu-24.04 and testbed-deploy-current-in-a-nutshell-ubuntu-24.04, which pin the OpenStack release rather than the OSISM one. The stable lanes pin manager_version to a released version and stay on ceph-ansible.

The testbed configuration adapts to a release past squid. set-ceph-version.sh, where the Ceph release enters a latest testbed's configuration, now does two things for any release other than quincy, reef and squid:

  • It disables the ceph-ansible container. On the latest track its tag follows ceph_version, and there is no ceph-ansible:tentacle. A nutshell lane leaves CEPH_STACK unset, so the cephadm switch on CEPH_STACK (Select the Ceph backend with CEPH_STACK #2960) does not reach it.
  • It removes rgw keystone api version from the Ceph environments. Tentacle removed the option, and writing it fails the cephadm configuration play. Reef and squid still default it to 2, so it stays for them.

These match the decisions cfg-cookiecutter makes for generated configurations, on the same release list.

Depends on

This lane deploys Tentacle on cephadm only with all of these in:

Part of

🤖 Generated with Claude Code

@ideaship
ideaship force-pushed the ceph-stack-selectable branch 2 times, most recently from 25cf7b5 to 8ba1984 Compare October 1, 2026 09:37
@ideaship
ideaship force-pushed the next-lane-cephadm-tentacle branch from b1ac71d to 57ffe73 Compare October 1, 2026 09:37
@ideaship ideaship self-assigned this Oct 1, 2026
@ideaship
ideaship force-pushed the ceph-stack-selectable branch from 8ba1984 to f6dd01c Compare October 1, 2026 13:32
@ideaship
ideaship force-pushed the next-lane-cephadm-tentacle branch from 57ffe73 to 2bc2d55 Compare October 1, 2026 13:33
@ideaship
ideaship marked this pull request as ready for review October 2, 2026 10:53
Base automatically changed from ceph-stack-selectable to main October 2, 2026 13:12
testbed-deploy-next-in-a-nutshell-ubuntu-24.04 is the lane for the next
OpenStack release, 2026.1, and it pinned Ceph to reef. reef is not the
Ceph this lane should be validating against 2026.1.

Point it at Tentacle, which osism/release now offers as a flavour for
latest. The tempest sibling inherits from this job and follows.

The Ceph backend is not set here. python-osism selects it from the OSISM
release: the latest track, which this lane runs, deploys with cephadm,
and ceph-ansible has no Tentacle path anyway. Naming a backend in the
job would only restate that, and the nutshell collection would ignore
it.

No other lane changes here, though the same release-keyed selection
moves every other latest nutshell lane to cephadm as well.

Assisted-by: Claude:claude-opus-5
Signed-off-by: Roger Luethi <luethi@osism.tech>
The next lane moves to Tentacle, and two things in the testbed
configuration assume a Ceph release ceph-ansible can deploy.

The manager starts a ceph-ansible container whose tag on the latest
track follows ceph_version, so a Tentacle testbed asks for
ceph-ansible:tentacle, which does not exist: ceph-ansible has no branch
past squid. This lane is a nutshell lane, where python-osism selects the
Ceph backend and CEPH_STACK stays unset, so the cephadm switch on
CEPH_STACK does not reach it.

The Ceph configuration sets "rgw keystone api version". Tentacle removed
that option together with Keystone v2.0 support, and writing a removed
option is not ignored: ceph config set answers EINVAL and the cephadm
configuration play fails. Reef and squid still default it to 2, so it
cannot simply be dropped.

set-ceph-version.sh is where the Ceph release enters a latest testbed's
configuration, so decide both there: for any release other than
quincy, reef and squid, disable the ceph-ansible container and remove
the option from the Ceph environments. This matches cfg-cookiecutter,
which makes the same two decisions on the same release list. Running
the script again changes nothing further.

Assisted-by: Claude:claude-opus-5-5
Signed-off-by: Roger Luethi <luethi@osism.tech>
@ideaship
ideaship force-pushed the next-lane-cephadm-tentacle branch from 2bc2d55 to 31c4d83 Compare October 2, 2026 14:01
@berendt
berendt merged commit 1d7a557 into main Oct 2, 2026
2 checks passed
@berendt
berendt deleted the next-lane-cephadm-tentacle branch October 2, 2026 17:45
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: Done

Development

Successfully merging this pull request may close these issues.

3 participants