Adds optional support for running the playbook on Synology DSM 7+, detected
automatically via /etc/synoinfo.conf so that non-Synology hosts are unaffected.
Includes DSM-native user/group management (synouser/synogroup), a requests
version constraint for Docker SDK compatibility, and a boot-fix service that
re-shares the volume mount and starts matrix services skipped by DSM's boot
ordering. The shared-mount volume path is configurable via
matrix_base_synology_volume_path, and the make-shared step only runs when the
volume is not already shared.
Co-authored-by: CKSit <sitchiuki@gmail.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a generic mechanism for exposing bridges' HTTP API (the provisioning
API, etc.) publicly on the Matrix domain, so tools like mautrix-manager
(https://github.com/mautrix/manager) can drive bridge login.
- Introduce global matrix_bridges_exposure_* vars (on by default),
exposing each supported bridge under `<matrix-fqn>/bridges/<bridge>`.
- mautrix-gmessages: make the provisioning shared secret configurable
(auto-generated in group_vars) so the provisioning API is enabled,
route the whole bridge HTTP port via Traefik, and populate
appservice.public_address.
Requests are authenticated by the bridge itself (per-user Matrix access
token for the provisioning API, homeserver token for the appservice
endpoints), not by the reverse proxy.
This is the first bridge converted; the other mautrix bridges will follow.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Tuwunel is a Matrix homeserver maintained by the matrix-construct
organisation. See https://matrix-construct.github.io/tuwunel/.
The rendered TOML emits only keys exposed as Ansible variables; the
rest fall back to tuwunel's upstream defaults. Anything not surfaced
can be set via the TUWUNEL_* env extension or by overriding the
template path.
Popular features Tuwunel adds variables for:
- OAuth2/OIDC identity providers (a list of `[[global.identity_provider]]`
blocks; brand-aware defaults for Google, GitHub, Keycloak, MAS, etc)
- LDAP and JWT authentication
- Media storage providers (native local and S3 with multipart upload)
- RocksDB tuning (compression, direct_io, parallelism, online backups)
- Native TLS dual-protocol mode
- Blurhashing, Sentry crash reporting
Auto-wired from existing playbook globals: well-known client URL,
TURN/coturn, MatrixRTC LiveKit URL, federation.
The `tuwunel-migrate-from-conduwuit` tag performs a binary-swap
migration. Migration from any other Conduit derivative is unsupported
and would corrupt the database.
Signed-off-by: Jason Volk <jason@zemos.net>
MAS now connects to the playbook-managed Postgres via a UNIX socket by
default (when available), matching the approach already used by Synapse.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The startup issue came from a timing dependency around coturn TLS certs:
- `matrix-coturn.service` depends on
`matrix-traefik-certs-dumper-wait-for-domain@<matrix-fqdn>.service`
- That waiter succeeds only after Traefik has obtained and dumped a cert for
the Matrix hostname (typically driven by homeserver labels/routes becoming
active)
- If coturn is started too early, it can block/fail waiting for cert files
that are not yet present
Historically, coturn priority was mode-dependent:
- `one-by-one`: coturn at 1500 (delayed after homeserver)
- other modes: coturn at 900 (before homeserver)
This could still trigger undesirable startup ordering and confusing behavior
in non-`one-by-one` modes, especially during initial bootstrap/restart flows
where cert availability lags service startup.
This change makes ordering explicit and consistent:
1. Introduce `matrix_homeserver_systemd_service_manager_priority` (default 1000)
in `roles/custom/matrix-base/defaults/main.yml`.
2. Use that variable for the homeserver service entry in
`group_vars/matrix_servers`.
3. Set coturn priority relative to homeserver priority in all modes:
`matrix_homeserver_systemd_service_manager_priority + 500`.
4. Update inline documentation comments in `group_vars/matrix_servers` to
match the new behavior and rationale.
Result:
- Homeserver/coturn ordering is deterministic and mode-agnostic.
- Coturn is intentionally started later than the homeserver by default,
reducing first-start certificate wait/fail races.
- Priority intent is now centralized and configurable via a dedicated
homeserver priority variable.
- Coturn may still be stated earlier, because the homeserver typically
has a `Wants` "dependency" on it, but that's alright
`ansible.builtin.yum` is no longer available in newer Ansible versions.
We may use `ansible.builtin.dnf` instead, but `ansible.builting.package`
seems like a safer bet that should work in both older & newer RHEL-based
distros.
Ref: https://github.com/ansible/ansible-lint/issues/4251
This is a backward-incompatible change. By default, Ansible creates
users with (e.g.) `/bin/sh` on Linux, so changing to a no shell
leads to different behavior.
That said, it appears that using a shell-less user works OK with regard
to Ansible execution and starting the systemd services/containers later on.
* continuwuity support
* continuwuity support
* use main instead of commit tag
* fix docker image link
* migration from conduwuit
* fix yaml lint
* backup directories and linter fixes
* linter fixes
* Add matrix_bridges_msc4190_enabled flag for using msc4190 on supported mautrix bridges.
* Apply to_json to msc4190 in mautrix configs
* Add | to_json to mautrix bridge registration io.element.msc4190.
* require matrix_synapse_experimental_features_msc3202_device_masquerading_enabled for matrix_bridges_msc4190_enabled
* Also add msc4190 support for mautrix-telegram
`matrix_container_global_registry_prefix_override` may look similar to
the old `matrix_container_global_registry_prefix` variable
(removed in d6bf789710), but it's different.
The old `matrix_container_global_registry_prefix` variable was just a hardcode of `docker.io/` and roles that
needed to refer to `docker.io/` could use it. However, this was:
- not used by all roles, because some need another registry (not `docker.io/`)
- used only by roles within the playbook (`roles/custom`), not external roles
Overriding the old `matrix_container_global_registry_prefix` variable was rather pointless,
as it didn't cover everything.
The new `matrix_container_global_registry_prefix_override` variable, on the other hand,
lets you override the registry prefix for all components, regardless of whether they
use `docker.io/` or another registry by default.
This is useful to people who have somehow mirrored all container images to their own registry,
as it provides them with a single variable they can flip to influence the whole playbook.
This is done for a few reasons:
- less globals and more indepdendence for each role is better. We rely
on various externally-hosted roles and they don't rely on this global
either.
- `matrix_container_global_registry_prefix` could make people think they
could just override this variable and have all their images pull from
elsewhere. This is rarely the case, unless you've taken special care
to mirror all the various components (from their respective
registries) to your own. In such a case, you probably know what you're
mirroring and can adjust individual variables.
- nowadays, various components live on different registries.
With Docker Inc tightening rate limits for Docker Hub, it's even more
likely that we'll see increased diversity in where images are hosted
cksit
Slavi Pantaleev
Suguru Hirahara
Jason Volk
QEDeD
Suguru Hirahara
The one with the braid
Benjamin Blacher
Suguru Hirahara
Virkkunen
Onestacked
Aine
László Várady