Recipe upgrade.
Commits on top of upstream main:
- a0e1579 chore: upgrade app to 26.7.0
Tested green on the cc-ci recipe CI server (full suite, cold, against this PR head). NOT merged — for operator review.
cc @trav @notplants
Bridge note: the ccci-bridge is silently dropping !testme (stale mounted Gitea-bot swarm secret → HTTP 401 "user does not exist"), so this run was driven directly via the Drone API with the bridge's exact query-string params (RECIPE=keycloak REF=526f6b5 PR=5 SRC=recipe-maintainers/keycloak) + a freshly-minted scoped GITEA_TOKEN for the runner's clone. The verdict was reflected back here manually as commit-status cc-ci/testme=failure. (Bridge secret not touched — operator infra.)
Diagnosis (this is NOT a recipe regression and NOT a stale test for the new version):
Both Drone runs fail identically at the install stage: deploy/readiness of the canonical baseline 10.8.0+26.6.3 (Keycloak 26.6.3 — the published tag immediately below the PR head's 10.8.1+26.6.4 label) times out with HTTP 404 on /realms/master over HTTPS. The upgrade stage (our 26.7.0) is never reached.
The 26.6.4 → 26.7.0 bump itself is verified green in a direct deploy on the cc-ci host (step 2b of the run): Keycloak 26.7.0 started in 7.6s, Migrating older model to 26.7.0, migrated realm master to 26.7.0, HTTP 200 on https://dev-keycloak.ci.commoninternet.net/realms/master. Torn down cleanly.
The canonical baseline 10.8.0+26.6.3 was ALSO redeployed directly on the host just now and serves HTTP 200 on /realms/master — so the recipe + the canonical tag + the host's abra are all fine.
The same PR head (526f6b5) was GREEN on 2026-07-24 (Drone build 1135) with an identical install baseline. The only delta between 07-24 GREEN and 08-03 RED is on the cc-ci/harness side: cc-ci main merged flake: bump nixpkgs + sops-nix (2026-08-03) (5366e061, includes 4772e312 nixpkgs bump + c541cb14 networking pin) earlier today. That infra change is the most likely root cause of the harness install readiness regression (the per-run Drone build 1179/1181 also did not persist a /var/lib/cc-ci-runs/1179 dir to the host, unlike concurrent other-recipe builds 1178/1180/1182 — suggesting a Drone-runner volume/env change).
A pre-existing recipe typo exists (compose.yml has depends_on: - mariadb referencing a non-existent service; the db service is named db and the volume is named mariadb), but it is not what is blocking CI: docker swarm docker stack deploy (what abra app deploy runs) ignores undefined depends_on, and the host deploy of both 26.6.3 and 26.7.0 succeeds with the typo present (HTTP 200). The harness's abra recipe config --images prepull step does rc=15 on this typo but treats it as a non-fatal skip ("deploy will pull as usual"). I did not edit it — it is out of scope for this upgrade PR and not the root cause.
Recommendation: this is a harness/infra regression on the 2026-08-03 nixpkgs bump, not a recipe/up graded-side problem. The recipe PR (#5) stands as-is (the 26.6.4 → 26.7.0 bump is correct and was direct-deploy-verified). Operator actions:
Investigate cc-ci main commits 4772e312/c541cb14/5366e061 (nixpkgs + networking pin) — likely changed Drone-runner abra/traefik/network behavior so the per-run install readiness 404s.
Re-run /recipe-upgrade keycloak --with-tests once the harness is fixed, to flip !testme to GREEN on this PR head (no recipe change expected to be needed).
(Optional, separate PR) fix the depends_on: - mariadb → - db typo for cleanliness.
NOT merged — for operator review. cc @trav @notplants
🌻 **cc-ci** — `keycloak` @ `526f6b5` ❌ **VERDICT=RED** (Drone builds 1179 + 1181, both red)
**Bridge note:** the `ccci-bridge` is silently dropping `!testme` (stale mounted Gitea-bot swarm secret → HTTP 401 "user does not exist"), so this run was driven **directly via the Drone API** with the bridge's exact query-string params (`RECIPE=keycloak REF=526f6b5 PR=5 SRC=recipe-maintainers/keycloak`) + a freshly-minted scoped GITEA_TOKEN for the runner's clone. The verdict was reflected back here manually as commit-status `cc-ci/testme=failure`. (Bridge secret not touched — operator infra.)
**Diagnosis (this is NOT a recipe regression and NOT a stale test for the new version):**
- Both Drone runs fail identically at the **install** stage: deploy/readiness of the canonical baseline `10.8.0+26.6.3` (Keycloak **26.6.3** — the published tag immediately below the PR head's `10.8.1+26.6.4` label) times out with HTTP 404 on `/realms/master` over HTTPS. The upgrade stage (our 26.7.0) is never reached.
- The 26.6.4 → 26.7.0 bump itself is **verified green in a direct deploy** on the cc-ci host (step 2b of the run): `Keycloak 26.7.0 started in 7.6s`, `Migrating older model to 26.7.0`, `migrated realm master to 26.7.0`, HTTP 200 on `https://dev-keycloak.ci.commoninternet.net/realms/master`. Torn down cleanly.
- The canonical baseline `10.8.0+26.6.3` was ALSO redeployed directly on the host just now and serves HTTP 200 on `/realms/master` — so the recipe + the canonical tag + the host's abra are all fine.
- The same PR head (526f6b5) was **GREEN on 2026-07-24** (Drone build 1135) with an identical install baseline. The only delta between 07-24 GREEN and 08-03 RED is on the cc-ci/harness side: cc-ci main merged `flake: bump nixpkgs + sops-nix (2026-08-03)` (`5366e061`, includes `4772e312` nixpkgs bump + `c541cb14` networking pin) earlier today. That infra change is the most likely root cause of the harness install readiness regression (the per-run Drone build 1179/1181 also did not persist a `/var/lib/cc-ci-runs/1179` dir to the host, unlike concurrent other-recipe builds 1178/1180/1182 — suggesting a Drone-runner volume/env change).
**A pre-existing recipe typo exists** (`compose.yml` has `depends_on: - mariadb` referencing a non-existent service; the db service is named `db` and the volume is named `mariadb`), but it is **not** what is blocking CI: docker swarm `docker stack deploy` (what `abra app deploy` runs) ignores undefined `depends_on`, and the host deploy of both 26.6.3 and 26.7.0 succeeds with the typo present (HTTP 200). The harness's `abra recipe config --images` prepull step does rc=15 on this typo but treats it as a non-fatal skip ("deploy will pull as usual"). I did **not** edit it — it is out of scope for this upgrade PR and not the root cause.
**Recommendation:** this is a harness/infra regression on the 2026-08-03 nixpkgs bump, not a recipe/up graded-side problem. The recipe PR (#5) stands as-is (the 26.6.4 → 26.7.0 bump is correct and was direct-deploy-verified). Operator actions:
1. Investigate cc-ci main commits `4772e312`/`c541cb14`/`5366e061` (nixpkgs + networking pin) — likely changed Drone-runner abra/traefik/network behavior so the per-run install readiness 404s.
2. Re-run `/recipe-upgrade keycloak --with-tests` once the harness is fixed, to flip `!testme` to GREEN on this PR head (no recipe change expected to be needed).
3. (Optional, separate PR) fix the `depends_on: - mariadb` → `- db` typo for cleanliness.
NOT merged — for operator review. cc @trav @notplants
Harness regression RESOLVED — verdict now GREEN (drone build #1199).
The canonical-baseline 404 that red'd builds #1179/#1181 during the 2026-08-03 weekly run is no longer reproducible: the same baseline install now converges and serves 200 through the full Drone path (router registered, artifacts persisted). The regression window was ~21:00 Aug 3 – early Aug 4 and healed after the cc-ci host's 2026-08-04 re-activation (bridge-secret deploy); the recipe was never at fault (as diagnosed — direct deploys were green throughout).
Status reflected as cc-ci/testme=success on head 526f6b57c1. Ready for operator review/merge.
**Harness regression RESOLVED — verdict now GREEN** (drone build [#1199](https://drone.ci.commoninternet.net/recipe-maintainers/cc-ci/1199)).
The canonical-baseline 404 that red'd builds #1179/#1181 during the 2026-08-03 weekly run is no longer reproducible: the same baseline install now converges and serves 200 through the full Drone path (router registered, artifacts persisted). The regression window was ~21:00 Aug 3 – early Aug 4 and healed after the cc-ci host's 2026-08-04 re-activation (bridge-secret deploy); the recipe was never at fault (as diagnosed — direct deploys were green throughout).
Status reflected as `cc-ci/testme=success` on head `526f6b57c1`. Ready for operator review/merge.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Recipe upgrade.
Commits on top of upstream main:
Tested green on the cc-ci recipe CI server (full suite, cold, against this PR head). NOT merged — for operator review.
cc @trav @notplants
!testme
🌻 cc-ci —
keycloak@526f6b57✅ passedfull logs · dashboard
!testme
🌻 cc-ci —
keycloak@526f6b57✅ passedfull logs · dashboard
🌻 cc-ci —
keycloak@526f6b5❌ VERDICT=RED (Drone builds 1179 + 1181, both red)Bridge note: the
ccci-bridgeis silently dropping!testme(stale mounted Gitea-bot swarm secret → HTTP 401 "user does not exist"), so this run was driven directly via the Drone API with the bridge's exact query-string params (RECIPE=keycloak REF=526f6b5 PR=5 SRC=recipe-maintainers/keycloak) + a freshly-minted scoped GITEA_TOKEN for the runner's clone. The verdict was reflected back here manually as commit-statuscc-ci/testme=failure. (Bridge secret not touched — operator infra.)Diagnosis (this is NOT a recipe regression and NOT a stale test for the new version):
10.8.0+26.6.3(Keycloak 26.6.3 — the published tag immediately below the PR head's10.8.1+26.6.4label) times out with HTTP 404 on/realms/masterover HTTPS. The upgrade stage (our 26.7.0) is never reached.Keycloak 26.7.0 started in 7.6s,Migrating older model to 26.7.0,migrated realm master to 26.7.0, HTTP 200 onhttps://dev-keycloak.ci.commoninternet.net/realms/master. Torn down cleanly.10.8.0+26.6.3was ALSO redeployed directly on the host just now and serves HTTP 200 on/realms/master— so the recipe + the canonical tag + the host's abra are all fine.526f6b5) was GREEN on 2026-07-24 (Drone build 1135) with an identical install baseline. The only delta between 07-24 GREEN and 08-03 RED is on the cc-ci/harness side: cc-ci main mergedflake: bump nixpkgs + sops-nix (2026-08-03)(5366e061, includes4772e312nixpkgs bump +c541cb14networking pin) earlier today. That infra change is the most likely root cause of the harness install readiness regression (the per-run Drone build 1179/1181 also did not persist a/var/lib/cc-ci-runs/1179dir to the host, unlike concurrent other-recipe builds 1178/1180/1182 — suggesting a Drone-runner volume/env change).A pre-existing recipe typo exists (
compose.ymlhasdepends_on: - mariadbreferencing a non-existent service; the db service is nameddband the volume is namedmariadb), but it is not what is blocking CI: docker swarmdocker stack deploy(whatabra app deployruns) ignores undefineddepends_on, and the host deploy of both 26.6.3 and 26.7.0 succeeds with the typo present (HTTP 200). The harness'sabra recipe config --imagesprepull step does rc=15 on this typo but treats it as a non-fatal skip ("deploy will pull as usual"). I did not edit it — it is out of scope for this upgrade PR and not the root cause.Recommendation: this is a harness/infra regression on the 2026-08-03 nixpkgs bump, not a recipe/up graded-side problem. The recipe PR (#5) stands as-is (the 26.6.4 → 26.7.0 bump is correct and was direct-deploy-verified). Operator actions:
4772e312/c541cb14/5366e061(nixpkgs + networking pin) — likely changed Drone-runner abra/traefik/network behavior so the per-run install readiness 404s./recipe-upgrade keycloak --with-testsonce the harness is fixed, to flip!testmeto GREEN on this PR head (no recipe change expected to be needed).depends_on: - mariadb→- dbtypo for cleanliness.NOT merged — for operator review. cc @trav @notplants
Harness regression RESOLVED — verdict now GREEN (drone build #1199).
The canonical-baseline 404 that red'd builds #1179/#1181 during the 2026-08-03 weekly run is no longer reproducible: the same baseline install now converges and serves 200 through the full Drone path (router registered, artifacts persisted). The regression window was ~21:00 Aug 3 – early Aug 4 and healed after the cc-ci host's 2026-08-04 re-activation (bridge-secret deploy); the recipe was never at fault (as diagnosed — direct deploys were green throughout).
Status reflected as
cc-ci/testme=successon head526f6b57c1. Ready for operator review/merge.Auto-closed by /recipe-upgrade: its changes are already in upstream main (merged upstream); mirror main re-synced
Pull request closed