`cron:set --global` wrote the property but emitted `unknown flag: --global` because the post-set `scheduler-cron-write` trigger received `--global` as the appName arg, which pflag rejected before reaching the trigger body.
The trigger args now omit appName for global writes, and the `scheduler-k3s` cron-write trigger short-circuits when called without an app since per-app reconciliation requires a real app name. The docker-local trigger already regenerates the global crontab from all apps so global `mailfrom`/`mailto` are picked up without any further changes.
The Docker daemon refuses any endpoint settings on the default bridge network and `docker compose` unconditionally attaches the service name as an alias on every joined network, so combining bridge with user-defined networks via compose's `networks:` block is impossible. When `vector-networks` is set, the compose template now joins only the configured networks; outbound to external sinks still works through user-defined network NAT. Additionally, `vector-image` and `vector-networks` are both global-only but `common.CommandPropertySet` silently accepts them at app level by merging global-only keys into the valid-property set, so both now reject explicitly in `validateSetValue`.
Adds a new global `vector-networks` property on the logs plugin that takes a comma-separated list of Docker networks. When set, the rendered compose file declares each network plus `bridge` as external and joins them on the vector service, so `docker compose up` reconciles attachments on every `logs:vector-start`. When unset, the existing `network_mode: bridge` template is preserved unchanged. The value is validated against `docker network inspect` at set time, rejects the reserved `bridge` entry, and is surfaced in `dokku logs:report` via `--logs-vector-global-networks`.
The ports plugin's `post-certs-update` trigger was rewriting every `https:443:*` mapping from the app's `http:80:*` mappings, silently overwriting any user-defined mapping such as `https:443:443` used by apps that terminate TLS inside the container. The trigger now skips the rewrite when an `https:443:*` mapping already exists, keeping the default behavior only when the app has no explicit HTTPS mapping configured.
Closes#8619.
The bare `tls-internal` key previously returned the computed value, so external tooling could not tell whether the property had been set on the app or was merely defaulting to `false`. The property is now also configurable with `--global`, the report exposes `computed-tls-internal` and `global-tls-internal` keys alongside the bare raw key, and the deploy path honors the per-app value with a fallback to the global value before the built-in default. Closes#8625.
Adds `certs-set` and `certs-remove` plugin triggers so other plugins can install or remove an app's SSL cert/key pair without shelling out to the `dokku certs:add` / `dokku certs:remove` subcommands. Shared implementations live as `fn-certs-set` and `fn-certs-remove` in `plugins/certs/internal-functions`, with the subcommands and the new triggers calling `verify_app_name` before delegating.
# History
## 0.38.4
Install/update via the bootstrap script:
```shell
wget -NP . https://dokku.com/install/v0.38.4/bootstrap.sh
sudo DOKKU_TAG=v0.38.4 bash bootstrap.sh
```
### Bug Fixes
- #8615: @josegonzalez Reject per-app sets for openresty global-only properties
- #8613: @josegonzalez Expose raw deploy-branch and keep-git-dir in git:report
- #8549: @josegonzalez Route CNB images through launcher on scheduler-k3s
### New Features
- #8614: @josegonzalez Split scheduler-docker-local report into raw, computed, and global
### Documentation
- #8603: @cheif Add `dokku-http-oauth` to community plugins
### Tests
- #8618: @josegonzalez Isolate scheduler-k3s registry tags per bats file
- #8616: @josegonzalez Migrate from junit_files to files in EnricoMi/publish-unit-test-result-action
- #8617: @josegonzalez Upgrade actions in shared build-image compose action
- #8609: @josegonzalez Skip packer lint job on dependabot PRs
- #8604: @dependabot[bot] chore(deps): bump python from 3.14.3-bookworm to 3.15.0b1-bookworm in /tests/apps/dockerfile-release
### Dependencies
- #8606: @dependabot[bot] chore(deps): bump golang.org/x/crypto from 0.50.0 to 0.51.0 in /plugins/common
- #8608: @dependabot[bot] chore(deps): bump github.com/traefik/traefik/v2 from 2.11.45 to 2.11.46 in /plugins/scheduler-k3s
- #8607: @dependabot[bot] chore(deps): bump dokku/openresty-docker-proxy from 0.10.0 to 0.11.0 in /plugins/openresty-vhosts
- #8605: @dependabot[bot] chore(deps): bump python from 3.14.3-alpine to 3.15.0b1-alpine in /docs/_build
`openresty:set <app>` previously accepted per-app writes for properties whose readers only consult the global store, so `:set myapp image foo` printed a success message while `:report myapp` kept showing the global default. The per-app form is now rejected with `The key '<key>' can only be set globally`, matching the behavior introduced for `caddy`, `haproxy`, and `traefik` in #8602.
The bare `init-process` and `parallel-schedule-count` keys previously returned the computed value, so external tooling could not tell whether a property had been set on the app or was merely defaulting. Both properties are now also configurable with `--global`, the report exposes `computed-*` and `global-*` keys alongside the bare raw keys, and the deploy path honors the global value before falling back to the linuxserver.io vendor heuristic.
The bare `deploy-branch` and `keep-git-dir` keys in `git:report` returned the computed (effective) value rather than the raw per-app value, with no separate `computed-*` key to distinguish "set per-app" from "falling back to global or default". This left external tooling unable to detect a per-app unset without out-of-band state. The bare keys now hold the raw per-app value (empty when unset) and new `computed-deploy-branch` and `computed-keep-git-dir` keys hold the effective value, matching the convention used by `nginx-vhosts`, `network`, and `builder`. Closes#8610.
The cron-id label could exceed Kubernetes' 63-byte cap when commands or
schedules were long, and an all-digit job-suffix or cron-id rendered as
an unquoted YAML scalar caused the API server to reject manifests. Run
pods built from dockerfiles also occasionally hit the 10s startup wait
on a cold image pull, even though the pod was scheduled correctly.
The cron-id is now stored as an annotation and a shorter hash is used as
the selector label. Every interpolated annotation and label value in the
cron-job and deployment templates is now quoted to prevent numeric
coercion, and the run-pod wait timeout is raised to 30 seconds.
After streaming logs from a run pod on scheduler-k3s, the apiserver may still report `PodRunning` for a short window due to kubelet status propagation lag, causing `dokku run` to fail with `Unable to attach as the pod is in an unknown state: Running`. Wait briefly for the pod to reach a terminal phase before classifying the outcome.
For short-lived commands the run pod can transition Running to Succeeded
between the running-pod check and the kubectl exec SPDY upgrade, leaving
the upgrade to fail with `container not found`. When stdout is not a TTY
(and DOKKU_FORCE_TTY is not set) the exec attach is only being used to
capture stdout, so stream the run pod logs in follow mode instead. The
pod's `TTLSecondsAfterFinished` of 60s keeps the kubelet's log file
readable for the duration of the call, eliminating the race.
Mirror the docker-local fix in #8525 for the k3s scheduler. CNB images default to a `/cnb/process/web` entrypoint that ignores incoming args, so non-web deployments, scheduled cron jobs, and ad-hoc `dokku run` / `cron:run` commands all need an explicit `launcher` entrypoint. The deployment and cron-job helm templates now set `command: [launcher]` when `image.type` is `pack`, and `TriggerSchedulerRun` sets the entrypoint to `launcher` for pack images while finishing the previously stubbed Procfile lookup branch so the resolved command is actually scheduled.
`caddy:set`, `haproxy:set`, and `traefik:set` previously accepted per-app writes for properties that only have a single host-wide reader, so `:set myapp image foo:bar` printed a success message while `:report myapp` kept showing the global default. The per-app form is now rejected with `The key '<key>' can only be set globally`, matching the existing rejection used for haproxy `refresh-conf` and traefik `challenge-mode`. Caddy `tls-internal` remains the only legitimate per-app property in this family.
The property name set via `app-json:set` is `appjson-path`, but the matching read-back flags on `app-json:report` were named `--app-json-selected`, `--app-json-global-selected`, and `--app-json-computed-selected`. The mismatch meant `dokku app-json:report <app> --app-json-appjson-path` (the form already documented in deployment-tasks.md) was rejected as an invalid flag, and the `--format json` output advertised keys that did not correspond to any settable property. The flags and JSON keys are renamed to `--app-json-appjson-path`, `--app-json-global-appjson-path`, and `--app-json-computed-appjson-path` so that the property name round-trips through set and report.
The `value_exists` variable in the report info-flag loop is no longer read after the `not deployed` failure was removed, and was already unused in domains, haproxy-vhosts, traefik-vhosts, caddy-vhosts, and openresty-vhosts. Drop the declaration and the trailing assignment so the loop reads cleanly across all plugins.
Several plugin `:report` subcommands erroneously failed with `not deployed` when an info-flag matched a property that was empty. Empty values are legitimate for configuration properties pre-deploy and the `--format json` path already returns them without error. Remove the `value_exists` check across nginx, checks, git, certs, scheduler-docker-local, and the builder-* plugins so the info-flag form behaves consistently with the JSON form.
The shipped catch-all default site uses `ssl_reject_handshake`, which is unsupported on nginx older than 1.19.4 and causes nginx to fail to start on Debian Bullseye. The postinst now detects the installed nginx version and installs an HTTP-only variant of the catch-all on older systems.
The bash and go validators previously each kept their own copy of the regex, which had to be updated in lockstep. Both bash wrappers now invoke the existing common binary via the same pattern as `verify_app_name`, leaving go as the single source of truth. The legacy `IsValidAppNameOld` rule is also widened to allow underscores again so apps created under the old naming rules can still be looked up through `VerifyAppName`'s either-rule fallback.
The previous app name validation regex permitted shell metacharacters such as `;`, `$`, backticks, `|`, and `&`. These names were embedded unquoted into the generated git pre-receive hook script, allowing an authenticated user to execute arbitrary commands as the dokku user simply by pushing to a remote with a crafted app name. App names are now restricted to lowercase alphanumerics, dots, and hyphens, and the hook script also quotes the app variable as a defense-in-depth measure.
Replaces the `DOKKU_ARCHIVE_MAX_SIZE` and `DOKKU_ARCHIVE_MAX_FILES` environment variables with global git properties (`archive-max-size` and `archive-max-files`), configurable via `dokku git:set --global` and surfaced through `dokku git:report --global`. Defaults remain `1073741824` bytes and `10000` entries.
Archives passed to git:from-archive and certs:add were extracted without symlink or path validation, allowing a crafted archive to write arbitrary files anywhere writable by the dokku user via symlink traversal. Extraction now pre-scans entries for absolute paths, parent traversal, and unsafe symlinks, applies the GNU tar `--no-unsafe-links` flag when available, and validates symlinks after extraction.
The previous use of `touch` before `netrc set` allowed the file to inherit the umask and be world-readable, exposing stored git credentials to local users. The set and unset paths now explicitly chmod 0600 and chown to the dokku user, and the plugin install hook repairs permissions on already-affected installations.
Add defense-in-depth sanitization for OpenResty include files to prevent
OS command injection via malicious filenames that break shell quoting in eval.
- Add filename validation in core-post-extract using regex [^a-zA-Z0-9_.-]
- Validate both http-includes and location-includes paths
- Abort deploy via dokku_log_fail on unsafe filenames
- Skip non-regular files (symlinks, directories) during extraction
- Add security regression test with unsafe filename containing space
- Keep existing guards in docker-args-process-deploy as belt-and-suspenders
- Update documentation to clarify allowed filename characters
Addresses CVSS 9.9 vulnerability where filenames like poc'$(cmd)'x.conf
could escape shell quoting and execute arbitrary commands during deploy.
The builder-dockerfile, builder-lambda, builder-nixpacks, builder-pack and builder-railpack `core-post-extract` triggers assigned `$2` to a local `SOURCECODE_WORK_DIR` but called `pushd "$TMP_WORK_DIR"`, which was unset. Bash 5.2 silently accepted `pushd ""`, so the bug stayed dormant. Bash 5.3 (shipped with Ubuntu 26.04) makes it a hard error and `set -e` aborts the trigger, causing every `git push` to fail with `pushd: null directory`.
