[adwk67]

Description

Part of https://github.com/stackabletech/internal-issues/issues/167

--- PASS: kuttl (210.31s)
    --- PASS: kuttl/harness (0.00s)
        --- PASS: kuttl/harness/znode_zookeeper-latest-3.9.6_openshift-false (26.52s)
        --- PASS: kuttl/harness/cluster-operation_zookeeper-latest-3.9.6_openshift-false (54.25s)
        --- PASS: kuttl/harness/logging_zookeeper-3.9.6_openshift-false (97.66s)
        --- PASS: kuttl/harness/smoke_zookeeper-3.9.6_use-server-tls-true_use-client-auth-tls-true_openshift-false (83.52s)
        --- PASS: kuttl/harness/delete-rolegroup_zookeeper-3.9.6_openshift-false (129.54s)
PASS

kuttl: the cluster-operation test has a new step 60-recreate-role-listener that deletes the role Listener and asserts that the discovery ConfigMap keeps its uid, data and resourceVersion once the Listener is re-created.

Manual checks on kind, with product images at stackableVersion: 26.3.0 vs. the CI image built from this PR 0.0.0-pr1105 :-

Scenario	Result
Fresh deploy	The discovery ConfigMap is written once, already with addresses; no version with empty ZOOKEEPER/ZOOKEEPER_HOSTS. Druid, which copies the ZooKeeper address from this ConfigMap into its own config, rolls out once (as on 26.3) instead of twice (as on main).
Role Listener deleted on a running cluster	The ConfigMap is unchanged (same uid and resourceVersion) while the Listener is re-created; no consumer rollouts.
Cluster paused before the Listener has an address	No reconcile errors while paused, no ConfigMap. Once unpaused, the ConfigMap is created with addresses straight away.
Upgrade 26.3.0 → PR-1105, role Listener present	Only the label change that every discovery ConfigMap gets on this upgrade; data unchanged, no empty version.
Upgrade 26.3.0 → PR-1105, role Listener missing (26.3.0 does not re-create a deleted one)	The stored ConfigMap is re-emitted before the new Listener has an address; data unchanged, no empty version (main writes empty values here).
Discovery ConfigMap deleted	Re-created at once, with addresses.
Discovery ConfigMap deleted while paused	Stays absent while paused; re-created with addresses on unpause.
Repeated reconciles	No ConfigMap change.
listenerClass change	The ConfigMap switches to the new address once the Listener publishes it, and back when reverted; no empty version.

What users notice

26.3.0 and 26.7.0 handle the role Listener the same way: the operator does not watch Listeners,
and the cluster reconcile fails while the Listener has no address. The differences below apply
to an upgrade from either release.

• No reconcile error while the role Listener has no address yet Before, the reconcile failed with "has no ingress addresses" (or a port-not-found error) and skipped the status update and orphan cleanup until the address appeared. Now it succeeds.
• The discovery ConfigMap content stays the same in that window. An existing ConfigMap keeps its values; a new one is created only once the address is known. Consumers never see empty ZOOKEEPER or ZOOKEEPER_HOSTS values. On a fresh deploy, Druid rolls out once, the same as on 26.3.0.
• The ConfigMap follows the address faster. Before, a change of the Listener's address only reached the ConfigMap on the next reconcile triggered by something else; now the ConfigMap is updated as soon as the Listener publishes the new address. Until then the old address is still advertised.
• A deleted role Listener is re-created. On 26.3.0 it stayed missing as no reconcile was triggered. Now it is re-created, and the discovery ConfigMap is left untouched meanwhile, so consumers do not roll out.
• While the cluster is paused, a discovery ConfigMap that does not exist yet is not created; it appears on unpause.
• On upgrade, the discovery ConfigMap gets one new resourceVersion, for a label change only; its data is unchanged.

Compared with the current main (unreleased), this PR removes the extra ConfigMap version with
empty ZOOKEEPER/ZOOKEEPER_HOSTS on every fresh deploy and every time the role Listener is
re-created, and with it the extra rollout of consumers such as Druid that copy the address into
their own config.

ZookeeperZnode discovery ConfigMaps are unchanged by this PR: a ZookeeperZnode still fails its
reconcile until the role Listener has an address.

Definition of Done Checklist

• Not all of these items are applicable to all PRs, the author should update this template to only leave the boxes in that are relevant
• Please make sure all these things are done and tick the boxes

Author

• Changes are OpenShift compatible
• CRD changes approved
• CRD documentation for all fields, following the style guide.
• Helm chart can be installed and deployed operator works
• Integration tests passed (for non trivial changes)
• Changes need to be "offline" compatible
• Links to generated (nightly) docs added
• Release note snippet added

Reviewer

• Code contains useful comments
• (Integration-)Test cases added
• Changelog updated
• Cargo.toml only contains references to git tags (not specific commits or branches)

Acceptance

• Feature Tracker has been updated
• Proper release label has been added
• Links to generated (nightly) docs added
• Release note snippet added
• Add type/deprecation label & add to the deprecation schedule
• Add type/experimental label & add to the experimental features tracker

[siegfriedweber]

LGTM

I have some comments that I would classify as "could have".