[DPE-10885] test(database): cover the client-relation events handler (7/7) - #222
Draft
marceloneppel wants to merge 6 commits into
Draft
[DPE-10885] test(database): cover the client-relation events handler (7/7)#222marceloneppel wants to merge 6 commits into
marceloneppel wants to merge 6 commits into
Conversation
marceloneppel
force-pushed
the
test/02-db-manager-lifecycle
branch
from
August 18, 2026 18:23
eb5e226 to
9dfff65
Compare
marceloneppel
force-pushed
the
test/02-db-events
branch
from
August 18, 2026 18:23
538d3da to
c1fab6a
Compare
This was referenced Aug 20, 2026
Draft
Drives real database-requested, relation-departed and relation-broken events through the handler rather than calling its methods, so the provider library's own leader gate and diff bookkeeping stay in the path under test. Covers both substrates' readiness guards, including the K8s removal guard that settles for a started member where a request also waits on the primary endpoint, and the K8s inline endpoint write the port kept. Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>
…andler The port moved the event.requested_entity_secret_content read out of the charm's _get_credentials try block into the DatabaseRequest construction, which left the manager's except ModelError guarding only a plain dict access it can never see. The vendored property resolves the secret on access and raises ModelError while a cross-model grant is still pending, so an ungranted requested-entity secret crashed the hook where the charm set a Blocked status and recovered on the next relation-changed. Restore the guard at the access boundary so the handler blocks instead of raising, matching both charms' 16/edge behavior. Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>
…_endpoints The ported test asserted the endpoints and read-only-endpoints values but never the URIs built from them, so building the read-write URI from the read-only endpoints survived the whole suite, and the original charm test's uris assertion was silently dropped in the port. Also restore the two other dropped scenarios: a relation without credentials publishes nothing, a prefix with no matching database clears the stale URIs, and a relation-scoped update leaves the other relations' databags untouched. Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>
The delete test never asserted remove_user_from_databases, which only fires when the deleted relation's database still lives in another relation's prefix mapping — the one arrange that reaches the loop, so dropping the loop survived the suite. Pin it with that arrange, along with the real secret-removal round trip in oversee_users, the rel_databases app-data cache key the Patroni pg_hba render reads, and the literal user-visible blocking messages, which were asserted through constants imported from the module under test and so could be reworded freely. Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>
The custom-username happy path in get_credentials — a granted secret returns its password verbatim, an ungranted one gets a generated password — was never executed by any test, and unblock_custom_user_errors had no coverage at all despite being the route update_unit_status takes for every custom-user blocking status. Cover its three outcomes (clear, invalid extra user roles elsewhere, forbidden or unreadable custom user) with literal status text. Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>
A few of the new assertions and arrange lines landed off the formatter's preference for line breaks; run it so tox -e lint passes at the branch tip. Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>
marceloneppel
force-pushed
the
test/02-db-manager-lifecycle
branch
from
August 20, 2026 12:34
9dfff65 to
445b125
Compare
marceloneppel
force-pushed
the
test/02-db-events
branch
from
August 20, 2026 12:34
335f89b to
3f548da
Compare
marceloneppel
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Issue
Last of the seven. PRs 5 and 6 cover the manager; the events handler's guards, defers and status mapping are still untested.
Solution
The request cases drive a real
database_requestedby writing to the requirer databag rather than calling the handler directly, so the provider library's own leader gate and diff bookkeeping stay in the path under test.Covers the happy path end to end, both defer paths (cluster not ready, peers not synced), the three blocking outcomes, the departing-unit flag, the leader and follower relation-broken paths, and the substrate split in the removal guard — K8s settles for a started member where VM also needs a primary. The K8s inline endpoint write is asserted present on K8s and absent on VM.
The PR also carries one fix the port review surfaced: building
DatabaseRequestreadevent.requested_entity_secret_contentoutside anyModelErrorguard, so an ungranted cross-model secret crashed the hook where both charms setMissing grant to requested entity secretand recovered on the next relation-changed. The guard now sits at the read itself, and its test drives the real event with the property raising. Landing it here rather than in 4/7 keeps the already-pushed branches below untouched.The mutation review also found holes in the manager tests from PRs 5 and 6, backfilled here: the published URIs and relation scoping of
update_endpoints, the delete path'sremove_user_from_databasesand secret cleanup, therel_databasescache key, the custom-username credentials paths,unblock_custom_user_errors, and the literal blocking messages (several tests had asserted constants imported from the module under test, which no test could reword-proof).Wiring up at:
Checklist