We have a flow like this:
1. matrix.DOMAIN vhost (matrix-domain.conf)
2. matrix-synapse vhost (matrix-synapse.conf); or matrix-corporal container, if enabled
3. (optional) matrix-synapse vhost (matrix-synapse.conf), if matrix-corporal enabled
4. matrix-synapse container
We are setting `X-Forwarded-For` correctly in step #1, but were
overwriting it in step #2 with something inaccurate.
Not doing anything in step #2 is better than doing the wrong thing.
It's probably best if we append another reverse-proxy address there
though, although what we're doing now (with this patch) seems to yield
the correct result (when matrix-corporal is not enabled).
When matrix-corporal is enabled, we still seem to do the wrong thing for
some reason. It's something to be fixed later on.
People who were disabling matrix-nginx-proxy (in favor of their own
nginx webserver) and also overriding `matrix_federation_public_port`,
found that the generated nginx configuration still hardcoded `8448`,
which forced their nginx server to use that, regardless of the fact
that `matrix_federation_public_port` was pointing elsewhere.
We now allow for the in-container federation port to be configurable,
and also automatically wire things properly.
We're talking about a webserver running on the same machine, which
imports the configuration files generated by the `matrix-nginx-proxy`
in the `/matrix/nginx-proxy/conf.d` directory.
Users who run an nginx webserver on some other machine will need to do
something different.
This give us the possibility to run multiple instances of
workers that that don't expose a port.
Right now, we don't support that, but in the future we could
run multiple `federation_sender` or `pusher` workers, without
them fighting over naming (previously, they'd all be named
something like `matrix-synapse-worker-pusher-0`, because
they'd all define `port` as `0`).
These are just defensive cleanup tasks that we run.
In the good case, there's nothing to kill or remove, so they trigger an
error like this:
> Error response from daemon: Cannot kill container: something: No such container: something
and:
> Error: No such container: something
People often ask us if this is a problem, so instead of always having to
answer with "no, this is to be expected", we'd rather eliminate it now
and make logs cleaner.
In the event that:
- a container is really stuck and needs cleanup using kill/rm
- and cleanup fails, and we fail to report it because of error
suppression (`2>/dev/null`)
.. we'd still get an error when launching ("container name already in use .."),
so it shouldn't be too hard to investigate.
This switches the `docker exec` method of spawning
Synapse workers inside the `matrix-synapse` container with
dedicated containers for each worker.
We also have dedicated systemd services for each worker,
so this are now:
- more consistent with everything else (we don't use systemd
instantiated services anywhere)
- we don't need the "parse systemd instance name into worker name +
port" part
- we don't need to keep track of PIDs manually
- we don't need jq (less depenendencies)
- workers dying would be restarted by systemd correctly, like any other
service
- `docker ps` shows each worker separately and we can observe resource
usage
We do this by creating one more layer of indirection.
First we reach some generic vhost handling matrix.DOMAIN.
A bunch of override rules are added there (capturing traffic to send to
ma1sd, etc). nginx-status and similar generic things also live there.
We then proxy to the homeserver on some other vhost (only Synapse being
available right now, but repointing this to Dendrite or other will be
possible in the future).
Then that homeserver-specific vhost does its thing to proxy to the
homeserver. It may or may not use workers, etc.
Without matrix-corporal, the flow is now:
1. matrix.DOMAIN (matrix-nginx-proxy/matrix-domain.conf)
2. matrix-nginx-proxy/matrix-synapse.conf
3. matrix-synapse
With matrix-corporal enabled, it becomes:
1. matrix.DOMAIN (matrix-nginx-proxy/matrix-domain.conf)
2. matrix-corporal
3. matrix-nginx-proxy/matrix-synapse.conf
4. matrix-synapse
(matrix-corporal gets injected at step 2).
I guess it didn't hurt to do it until now, but it's not great serving
federation APIs on the client-server API port, etc.
matrix-corporal doesn't work yet (still something to be solved in the
future), but its firewalling operations will also be sabotaged
by Client-Server APIs being served on the federation port (it's a way to get around its firewalling).
While administering we will occasionally invoke this script interactively with the "non-interactive" switch still there, yet still sit at the desk waiting for 300 seconds for this timer to run out.
The systemd-timer already uses a 3h randomized delay for automatic renewals, which serves this purpose well.
Fixes https://github.com/spantaleev/matrix-docker-ansible-deploy/issues/756
Related to https://github.com/spantaleev/matrix-docker-ansible-deploy/issues/737
I feel like timers are somewhat more complicated and dirty (compared to
cronjobs), but they come with these benefits:
- log output goes to journald
- on newer systemd distros, you can see when the timer fired, when it
will fire, etc.
- we don't need to rely on cron (reducing our dependencies to just
systemd + Docker)
Cronjobs work well, but it's one more dependency that needs to be
installed. We were even asking people to install it manually
(in `docs/prerequisites.md`), which could have gone unnoticed.
Once in a while someone says "my SSL certificates didn't renew"
and it's likely because they forgot to install a cron daemon.
Switching to systemd timers means that installation is simpler
and more unified.
Not sure if it breaks with them or not, but no other directive
uses quotes and the nginx docs show examples without quotes,
so we're being consistent with all of that.
The different configurations are now all lower case, for consistent
naming.
`matrix_nginx_proxy_ssl_config` is now called
`matrix_nginx_proxy_ssl_preset`. The different options for "modern",
"intermediate" and "old" are stored in the main.yml file, instead of
being hardcoded in the configuration files. This will improve the
maintainability of the code.
The "custom" preset was removed. Now if one of the variables is set, it
will use it instead of the preset. This will allow to mix and match more
easily, for example using all the intermediate options but only
supporting TLSv1.2. This will also provide better backward
compatibility.
A new variable called `matrix_nginx_proxy_ssl_config` is created for
configuring how the nginx proxy configures SSL. Also a new configuration
validation option and other auxiliary variables are created.
A new variable configuration called `matrix_nginx_proxy_ssl_config` is
created. This allow to set the SSL configuration easily using the
default options proposed by Mozilla. The default configuration is set to
"Intermediate", removing the weak ciphers used in the old
configurations.
The new variable can also be set to "Custom" for a more granular control.
This allows to set another three variables called:
- `matrix_nginx_proxy_ssl_protocols`,
- `matrix_nginx_proxy_ssl_prefer_server_ciphers`
- `matrix_nginx_proxy_ssl_ciphers`
Also a new task is added to validate the SSL configuration variable.
The Docker 19.04 -> 20.10 upgrade contains the following change
in `/usr/lib/systemd/system/docker.service`:
```
-BindsTo=containerd.service
-After=network-online.target firewalld.service containerd.service
+After=network-online.target firewalld.service containerd.service multi-user.target
-Requires=docker.socket
+Requires=docker.socket containerd.service
Wants=network-online.target
```
The `multi-user.target` requirement in `After` seems to be in conflict
with our `WantedBy=multi-user.target` and `After=docker.service` /
`Requires=docker.service` definitions, causing the following error on
startup for all of our systemd services:
> Job matrix-synapse.service/start deleted to break ordering cycle starting with multi-user.target/start
A workaround which appears to work is to add `DefaultDependencies=no`
to all of our services.
Since the switch from `-v` to `--mount` (in 1fca917ad1),
we've regressed when `matrix_ssl_retrieval_method == 'none'`.
In such a case, we don't create `/matrix/ssl` directories at all
and shouldn't be trying to mount them into the `matrix-nginx-proxy`
container.
Previously, with `-v`, Docker would auto-create them, effectively hiding
our mistake. Now that `--mount` doesn't do such auto-creation magic,
the `matrix-nginx-proxy` container was failing to start.
Fixes https://github.com/spantaleev/matrix-docker-ansible-deploy/issues/734
`-v` magically creates the source destination as a directory,
if it doesn't exist already. We'd like to avoid this magic
and the potential breakage that it might cause.
We'd rather fail while Docker tries to find things to `--mount`
than have it automatically create directories and fail anyway,
while having contaminated the filesystem.
There's a lot more `-v` instances remaining to be fixed later on.
This is just some start.
Things like `matrix_synapse_container_additional_volumes` and
`matrix_nginx_proxy_container_additional_volumes` were not changed to
use `--mount`, as options for each one are passed differently
(`ro` is `ro`, but `rw` doesn't exist and `slave` is `bind-propagation=slave`).
To avoid breaking people's custom volume mounts, we keep it as it is for now.
A deficiency with `--mount` is that it lacks the `z` option (SELinux
ownership changes), and some of our `-v` instances use that. I'm not
sure how supported SELinux is for us right now, but it might be,
and breaking that would not be a good idea.
The regex introduced in 63a49bb2dc seems to take precedence
over the bare location blocks, causing a regression.
> It is important to understand that, by default, Nginx will serve regular expression matches in preference to prefix matches.
> However, it evaluates prefix locations first, allowing for the administer to override this tendency by specifying locations using the = and ^~ modifiers.
Source: https://www.digitalocean.com/community/tutorials/understanding-nginx-server-and-location-block-selection-algorithms
In #628 I proposed a CORS change that turns out not to be the root of the issue. Caffeine-addled diagnosis leads to sloppy thinking, and this change should be reverted. In fact, if left it will cause problems for new installations.
Even with the v2 updates listed in #503 and partially addressed in #614, this is still needed to enable identity services to function with Element Desktop/Web. Testing on multiple clients with a clean config has confirmed this, at least for my installation.
Fix regression since 2a50b8b6bb (#597).
Dimension is intended to be embedded in various clients,
be it the Element service that we host (at element.DOMAIN),
some other Element (element-desktop running locally), etc.
Depending on the distro, common commands like sleep and chown may either
be located in /bin or /usr/bin.
Systemd added path lookup to ExecStart in v239, allowing only the
command name to be put in unit files and not the full path as
historically required. At least Ubuntu 18.04 LTS is however still on
v237 so we should maintain portability for a while longer.
· 😅 How to keep this in sync with the matrix-synapse documentation?
· regex location matching is expensive
· nginx syntax limit: one location only per block / statement
· thus, lots of duplicate statements in this file
Continuation of #234 (Github Pull Request).
I had unintentionally updated the documentation for the feature,
saying the page is available at `https://matrix.DOMAIN/nginx_status`.
Looks like it wasn't the case, going against my expectations.
I'm correcting this with this patch.
The status page is being made available on both HTTP and HTTPS.
Serving over HTTP is likely necessary for services like
Longview
(https://www.linode.com/docs/platform/longview/longview-app-for-nginx/)
Reasoning is the same as for matrix-org/synapse#5023.
For us, the journal used to contain `docker` for all services, which
is not very helpful when looking at them all together (`journalctl -f`).
`matrix_nginx_proxy_data_path` has always served as a base path,
so we're renaming it to reflect that.
Along with this, we're also introducing a new "data path" variable
(`matrix_nginx_proxy_data_path`), which is really a data path this time.
It's used for storing additional, non-configuration, files related to
matrix-nginx-proxy.
If someone decides to not use `/.well-known/matrix/server` and only
relies on SRV records, then they would need to serve tcp/8448 using
a certificate for the base domain (not for the matrix) domain.
Until now, they could do that by giving the certificate to Synapse
and setting it terminate TLS. That makes swapping certificates
more annoying (Synapse requires a restart to re-read certificates),
so it's better if we can support it via matrix-nginx-proxy.
Mounting certificates (or any other file) into the matrix-nginx-proxy container
can be done with `matrix_nginx_proxy_container_additional_volumes`,
introduced in 96afbbb5a.
Certain use-cases may require that people mount additional files
into the matrix-nginx-proxy container. Similarly to how we do it
for Synapse, we are introducing a new variable that makes this
possible (`matrix_nginx_proxy_container_additional_volumes`).
This makes the htpasswd file for Synapse Metrics (introduced in #86,
Github Pull Request) to also perform mounting using this new mechanism.
Hopefully, for such an "extension", keeping htpasswd file-creation and
volume definition in the same place (the tasks file) is better.
All other major volumes' mounting mechanism remains the same (explicit
mounting).
Continuation of 1f0cc92b33.
As an explanation for the problem:
when saying `localhost` on the host, it sometimes gets resolved to `::1`
and sometimes to `127.0.0.1`. On the unfortunate occassions that
it gets resolved to `::1`, the container won't be able to serve the
request, because Docker containers don't have IPv6 enabled by default.
To avoid this problem, we simply prevent any lookups from happening
and explicitly use `127.0.0.1`.
We run containers as a non-root user (no effective capabilities).
Still, if a setuid binary is available in a container image, it could
potentially be used to give the user the default capabilities that the
container was started with. For Docker, the default set currently is:
- "CAP_CHOWN"
- "CAP_DAC_OVERRIDE"
- "CAP_FSETID"
- "CAP_FOWNER"
- "CAP_MKNOD"
- "CAP_NET_RAW"
- "CAP_SETGID"
- "CAP_SETUID"
- "CAP_SETFCAP"
- "CAP_SETPCAP"
- "CAP_NET_BIND_SERVICE"
- "CAP_SYS_CHROOT"
- "CAP_KILL"
- "CAP_AUDIT_WRITE"
We'd rather prevent such a potential escalation by dropping ALL
capabilities.
The problem is nicely explained here: https://github.com/projectatomic/atomic-site/issues/203
This makes all containers (except mautrix-telegram and
mautrix-whatsapp), start as a non-root user.
We do this, because we don't trust some of the images.
In any case, we'd rather not trust ALL images and avoid giving
`root` access at all. We can't be sure they would drop privileges
or what they might do before they do it.
Because Postfix doesn't support running as non-root,
it had to be replaced by an Exim mail server.
The matrix-nginx-proxy nginx container image is patched up
(by replacing its main configuration) so that it can work as non-root.
It seems like there's no other good image that we can use and that is up-to-date
(https://hub.docker.com/r/nginxinc/nginx-unprivileged is outdated).
Likewise for riot-web (https://hub.docker.com/r/bubuntux/riot-web/),
we patch it up ourselves when starting (replacing the main nginx
configuration).
Ideally, it would be fixed upstream so we can simplify.
The matrix-nginx-proxy role can now be used independently.
This makes it consistent with all other roles, with
the `matrix-base` role remaining as their only dependency.
Separating matrix-nginx-proxy was relatively straightforward, with
the exception of the Mautrix Telegram reverse-proxying configuration.
Mautrix Telegram, being an extension/bridge, does not feel important enough
to justify its own special handling in matrix-nginx-proxy.
Thus, we've introduced the concept of "additional configuration blocks"
(`matrix_nginx_proxy_proxy_matrix_additional_server_configuration_blocks`),
where any module can register its own custom nginx server blocks.
For such dynamic registration to work, the order of role execution
becomes important. To make it possible for each module participating
in dynamic registration to verify that the order of execution is
correct, we've also introduced a `matrix_nginx_proxy_role_executed`
variable.
It should be noted that this doesn't make the matrix-synapse role
dependent on matrix-nginx-proxy. It's optional runtime detection
and registration, and it only happens in the matrix-synapse role
when `matrix_mautrix_telegram_enabled: true`.
With this change, the following roles are now only dependent
on the minimal `matrix-base` role:
- `matrix-corporal`
- `matrix-coturn`
- `matrix-mailer`
- `matrix-mxisd`
- `matrix-postgres`
- `matrix-riot-web`
- `matrix-synapse`
The `matrix-nginx-proxy` role still does too much and remains
dependent on the others.
Wiring up the various (now-independent) roles happens
via a glue variables file (`group_vars/matrix-servers`).
It's triggered for all hosts in the `matrix-servers` group.
According to Ansible's rules of priority, we have the following
chain of inclusion/overriding now:
- role defaults (mostly empty or good for independent usage)
- playbook glue variables (`group_vars/matrix-servers`)
- inventory host variables (`inventory/host_vars/matrix.<your-domain>`)
All roles default to enabling their main component
(e.g. `matrix_mxisd_enabled: true`, `matrix_riot_web_enabled: true`).
Reasoning: if a role is included in a playbook (especially separately,
in another playbook), it should "work" by default.
Our playbook disables some of those if they are not generally useful
(e.g. `matrix_corporal_enabled: false`).
As suggested in #63 (Github issue), splitting the
playbook's logic into multiple roles will be beneficial for
maintainability.
This patch realizes this split. Still, some components
affect others, so the roles are not really independent of one
another. For example:
- disabling mxisd (`matrix_mxisd_enabled: false`), causes Synapse
and riot-web to reconfigure themselves with other (public)
Identity servers.
- enabling matrix-corporal (`matrix_corporal_enabled: true`) affects
how reverse-proxying (by `matrix-nginx-proxy`) is done, in order to
put matrix-corporal's gateway server in front of Synapse
We may be able to move away from such dependencies in the future,
at the expense of a more complicated manual configuration, but
it's probably not worth sacrificing the convenience we have now.
As part of this work, the way we do "start components" has been
redone now to use a loop, as suggested in #65 (Github issue).
This should make restarting faster and more reliable.
teutat3s
