Normally, Home Assistant uses a SQLite database for storing state
history. On a Raspberry Pi with only an SD card for storage like
*hass1.pyrocufflink.blue*, this can become extremely slow, especially
for large data sets. To speed up features like history and logbook,
Home Assistant supports using an external database engine such as
PostgreSQL or MariaDB.
The *hassdb* role and corresponding `hassdb.yml` playbook deploys a
PostgreSQL server for Home Assistant to use. It needs only to create
the role and database, as Home Assistant manages its own schema.
The *postgresql-setup* service is no longer necessary, as upstream has
fixed the SELinux policy to allow root to invoke the `postgresql-setup`
command directly.
This commit adds a task to generate a PostgreSQL configuration file from
a template. Previously, the default configuration file generated by
`initdb` was sufficient, but in order to enable SSL connections, some
changes to it are required.
Naturally, SSL connections require a server certificate, so the
*postgresql-server* role will now also copy certificate files to the
managed node, if any.
Fedora has renamed the *strongswan* service to *strongswan-starter*.
The *strongswan* service now controls strongSwan via Vici, which uses a
different configuration format and is not compatible with the files in
`/etc/strongswan/ipsec.d`. As I am migrating everything to Wireguard
now, it does not make sense to rewrite all of the IPsec configuration in
this new format, so using the legacy format with the renamed service
makes more sense.
Because the Home Assistant user's home directory is on `/var`, Python
packages installed in the "user site" do not get the correct SELinux
labels and thus run in the wrong domain. This causes a lot of AVC
denials and other issues that prevent Home Assistant from working
correctly.
To resolve this issue, Home Assistant is now installed in a virtual
environment at `/usr/local/homeassistant`. This directory is still
owned by the Home Assistant user, allowing Home Assistant to manage
packages installed there. Since it is rooted under `/usr`, files are
labelled correctly and processes launched from executables there will
run in the correct domain.
The main network, *pyrocufflink.blue* (172.30.0.0/26) is now on VLAN 1
instead of VLAN 30. This changed when I replaced the Cisco SG200-26
with the UniFI Switch 48, to simplify configuration of all of the
Ubiquiti devices.
For some time, I have been trying to design a new configuration for the
reverse proxy on port 443 to correctly handle all the types of traffic
on that port. In the original implementation, all traffic on port 443
was forwarded by the gateway to HAProxy. HAproxy then used TLS SNI to
route connections to the correct backend server based the requested host
name. This allowed both HTTPS and OpenVPN-over-TLS to use the same
port, however it was not without issues. A layer 4 (TCP) proxy like
this "hides" the real source address of clients connecting to the
backend, which makes IP-based security (e.g. rate limiting, blacklists,
etc.) impossible at the application level. In particular, Nextcloud,
which implements rate limiting was constantly imposing login delays on
all users, because legitimate traffic was indistinguishable from
Internet background noise.
To alleviate these issues, I needed to change the proxy to operate in
layer 7 (HTTP) mode, so that headers like *X-Forwarded-For* and
*X-Forwarded-Host* could be added. Unfortunately, this was not easy,
because of the simultaneous requirement to forward OpenVPN traffic.
HAProxy can only do SNI inspection in TCP mode. So, I began looking for
an alternate way to proxy both HTTP and non-HTTP traffic on the same
port.
The HTTP protocol defines the `CONNECT` method, which is used by forward
proxies to tunnel HTTPS over plain HTTP. OpenVPN clients support
tunneling OpenVPN over HTTP using this method as well. HAProxy has
limited support for the CONNECT method (i.e. it doesn't do DNS
resolution, and I could find no way of restricting the destination) with
the `http_proxy` option, so I looked for alternate proxy servers that
had more complete support. Unsurprisingly, Apache HTTPD has the most
complete implementation of the `CONNECT` method (Nginx doesn't support
it at all). Using a name-based virtual host on port 443, Apache will
accept requests for *vpn.pyrocufflink.net* (using TLS SNI) and allow the
clients to use the `CONNECT` method to create a tunnel to the OpenVPN
server. This requires OpenVPN clients to a) use *stunnel* to wrap plain
HTTP proxy connections in TLS and b) configure OpenVPN to use the
TLS-wrapped HTTP proxy.
With Apache accepting all incoming connections, it was trivial to also
configure it as a layer 7 forward proxy for Bitwarden, Gitea, Jenkins,
and Nextcloud. Unfortunately, proxying for the other websites
(darkchestofwonders.us, chmod777.sh, dustin.hatch.name) was not quite as
straightforward. These websites would need to have an internal name
that differed from their external name, and thus a certificate valid for
that name. Rather than reconfigure all of these sites and set all of
that up, I decided to just move the responsibility for handling direct
connections from outside to the *web0* and eliminate the dedicated
reverse proxy. This was not possible before, because Apache could not
forward the OpenVPN traffic directly, but now with the forward proxy
configuration, there is no reason to have a separate server for these
connections.
Overall, I am pleased with how this turned out. It makes the OpenVPN
configuration simpler (*stunnel* no longer needs to run on the OpenVPN
server itself, since Apache is handling TLS termination), eliminates a
network hop for the websites, makes the reverse proxy configuration for
the other web applications much easier to understand, and resolves the
original issue of losing client connection information.
A name-based HTTP (not HTTPS) virtual host for *pyrocufflink.net* is
necessary to ensure requests are handled properly, now that there is
another HTTP virtual host (chmod777.sh) defined on the same server.
This commit updates the configuration for *pyrocufflink.net* to use the
wildcard certificate managed by *lego* instead of an unique certificate
managed by *certbot*.
*chmod777.sh* is a simple static website, generated by Hugo. It is
built and published from a Jenkins pipeline, which runs automatically
when new commits are pushed to Gitea.
The HTTPS certificate for this site is signed by Let's Encrypt and
managed by `lego` in the `certs` submodule.
This commit adds front-end and back-end configuration for HAProxy to
proxy HTTP/HTTPS for
*nextcloud.pyrocufflink.net*/*nextcloud.pyrocufflink.blue* to
*cloud0.pyrocufflink.blue*.
The *nextcloud* role installs Nextcloud from the specified release
archive, downloading it to the control machine first if necessary, and
configures Apache and PHP-FPM to serve it.
The `nextcloud.yml` playbook uses the *cert* role to install the X.509
certificate for the Nextcloud server, sets up Apache HTTPD with the
*apache* role, and installs Nextcloud using the *nextcloud* role.
The host *cloud0.pyrocufflink.blue* is the Nextcloud server for
Pyrocufflink.
The *cert* role is intended to be a generic, reusable role to copy an
X.509 certificate and/or private key file to managed nodes. It is
intended to be included in a playbook with at least the `cert_src` and
`cert_dest` variables defined, e.g.:
```
- hosts: whatever
roles:
- role: cert
cert_src: whatever.cer
cert_dest: /path/to/whatever.cer
```
the `haproxy_ssl_default_bind_options` variable is not defined for
machines running Fedora, because this parameter is not used in the
default configuration file there.
I seem to have forgotten how I got the RPM for Gitea. I think I built
it, but I cannot find the spec file, nor the RPM package. Since this is
clearly not reproducible, I decided to switch to using the binary
provided by upstream for now, until either I or Fedora get around to
making a better RPM.
Installing Gitea from the upstream binary is simple: just download it
and copy it to `/usr/local/bin`. Of course, the OS user and systemd
unit have to be managed by configuration policy when it's installed this
way.
*burp1.pyrocufflink.blue* will replace *burp0.pyrocufflink.blue* as the
BURP server for Pyrocufflink. It is a physical machine (Fitlet), making
it simpler to manage the USB drives. The old virtual machine will be
decommissioned soon.
Ansible replaced the `version_compare` filter with a `version_compare`
test that does the same thing. The former is completely gone now,
causing the template to fail to render, so its usage of that filter
needs to be updated.
Using the generic *burp.pyrocufflink.blue* name will allow easier
transition to a new BURP server. However, since this is not the actual
name, it cannot be used for task delegation, so a separate variable is
required to store the real name of the BURP server. This is only used
during client deployment, and not by BURP itself.
The `Alias` configuration for Certbot needs to be configured before any
other locations, to ensure the `/.well-known` path is always served from
the local filesystem. If another drop-in configuration file (e.g.
`bitwarden.conf`) is ordered before it, it may override this
configuration and prevent Let's Encrypt from working.
In order to allow Jenkins to connect to the Docker daemon socket, the
socket must be owned by the *docker* group, and the *jenkins* user must
be a member of it.
This commit adds an HAProxy backend for Bitwarden, and adds ACL rules to
the frontend to proxy traffic to *bitwarden.pyrocufflink.blue* or
*bitwarden.pyrocufflink.net* to it.
Since the same certificate is used for LDAPS and RADIUS (EAP-TLS), it
makes more sense to store it only once, with the later file as a symlink
to the former.
This commit configures *bw0.pyrocufflink.blue* as a BURP client, so that
the Bitwarden data can be backed up. A pre-backup script is used to
take a consistent snapshot of the SQLite database before copying it to
the BURP server.
The BURP server runs as user *burp*, and nas such, requires that the
client-specific configuration files be owned by that user so they can be
read when a client connects.
Newer versions of the BURP client require `status_port` to be set. This
commit updates the `burp.conf.j2` template to more closely match the
default configuration shipped with the *burp* package, including setting
this new value.
Newer versions of Gitea need a JWT secret for Oauth2. Gitea will
attempt to generate one at startup if it is not already specified in the
configuration file, but this will fail since the file is not writable by
the user running the service. As such, it must be set via configuration
policy.
The point of the "wheel host" is to serve as a repository of Python
packages (wheels) built by Jenkins for consumption by `pip` et al. For
applications and libraries that do not provide all of their dependencies
as binary packages, this makes a convenient way to install them without
requiring all of the build tools and dependencies on the destination
machine.
The idea here is that a Jenkins job runs `pip wheel` for a distribution
package name or `requirements.txt` file and then uploads the resulting
wheel files using `rsync`. Apache is configured to serve the upload
directory with an index compatible with `pip`'s `--find-links`.
The *hass-dhcp* role installs dnsmasq and configures it to serve DHCP
requests on the Home Assistant network. Since this network is not
routed, the regular DHCP relay/server setup will not work.
This commit adds a systemd unit to enable the Kernel Same-page Merging
daemon on VM hosts. This allows much greater virtual machine density,
especially when many VMs are running the same guest OS.
Debian does not support system-wide SSL cipher suite profiles of course,
so these options need to be specified explicitly when deploying HAProxy
on Debian-based machines.
This commit updates the net-ifaces scripts for both *vmhost0* and
*vmhost1* to create VLAN and bridge interfaces for the Management and
Home Assistant networks.
Dustin