Filesystems like NFS and CIFS require "helper" utilities (i.e.
`mount.nfs` and `mount.cifs`, respectively). These need to be installed
in order for a system to be able to mount those filesystems.
The current shared storage system uses NFSv4, and as such, the
*nfs-utils* package needs to be installed on the VM hosts.
With the transition away from *dhcpcd* on the VM hosts, there is no
longer any need for a custom wait script that must run prior to
attempting to mount the shared filesystem. This dramatically simplifies
the configuration necessary for shared storage.
I don't really see any reason why the shared storage configuration needs
to be managed by a separate role. The *vmhost* role is not really
generic anyway, and will probably not work for any other VM host
deployment besides the two machines running now. As such, I think it
makes sense to move the task to mount the shared filesystem into the
*vmhost* role and drop the *dch-storage-net* role.
The *libvirt-daemon-driver-network* package provides support for
managing virtual networks with libvirt. It is necessary in order to use
managed networks in VM configuration, as opposed to directly specifying
VM network interfaces in their domain configuration.
*vmhost0.pyrocufflink.blue* no longer uses `dhcpcd` for network
configuration, but *systemd-networkd*.
The host-specific network settings for a VM host include the
configuration for the management interface, as well as the configuration
of the physical ports that make up the bonded interfaces.
Originally, the network configuration for the VM networks and the
storage network was configured using the *netifaces* role. This has
effectively stopped working in recent versions of Fedora, as it sort of
relied on `dhcpcd`, which has not been maintained in Fedora for a while
and no longer behaves correctly. After evaluating *NetworkManager* as a
replacement, I decided that *systemd-networkd* is a more appropriate
solution.
There are effectively two "layers" of network configuration needed for
the VM hosts: the host-specific settings, and the common settings. The
host-specific settings include such properties as the IP address of the
management interface and the names of the physical ports that make up
the bonded interfaces. The common settings are the bonded interfaces,
the VLAN interfaces created on top of the bond, and the bridges that
provide access to VMs.
To configure the host-specific settings, each host simply needs the
appropriate `networkd_*` variables in its `host_vars` file. For the
common settings, we apply the *systemd-networkd* role again in the
`vmhost.yml` with different values for these variables. Thus,
effectively, `systemd-networkd.yml` manages the host-specific settings,
while `vmhost.yml` manages the common settings.
*systemd-networkd* is (currently) my preferred way to manage network
interfaces on machines running Fedora. The *systemd-networkd* role
provides a generic way to configure network links, devices, and
interfaces, using Ansible variables to generate network unit
configuration files.
*hass1.pyrocufflink.blue* and *hassdb0.pyrocufflink.blue* were part of
the old Home Assistant deployment. Everything has been migrated to
*hass2.pyrocufflink.blue*, so these machines can be decommissioned now.
I couldn't get RTMP to work on the Back Yard camera because the `ffmpeg`
process kept crashing:
```
ffmpeg.back_yard.clips_rtmp ERROR : av_interleaved_write_frame(): Connection reset by peer
ffmpeg.back_yard.clips_rtmp ERROR : [flv @ 0x5562090c8ec0] Failed to update header with correct duration.
ffmpeg.back_yard.clips_rtmp ERROR : [flv @ 0x5562090c8ec0] Failed to update header with correct filesize.
ffmpeg.back_yard.clips_rtmp ERROR : Error writing trailer of rtmp://127.0.0.1/live/back_yard: Connection reset by peer
watchdog.back_yard INFO : Terminating the existing ffmpeg process...
watchdog.back_yard INFO : Waiting for ffmpeg to exit gracefully...
```
I thought increasing the value of `--shm-size` argument for `podman`
would help, but even going as high as 1024 mebibytes did not resolve the
problem.
Ultimately, I decided that it is not really necessary to view the full
4k stream in real time. The back yard camera supports three streams, so
I set them all up for different roles. I briefly considered using a
single 1080p stream for both object detection and RTMP streaming, but
this consumed considerable CPU time, so I decided against it for now. I
may re-evaluate that option if I decide to purchase a TPU.
The `collectd_df` variable can be used to configure the *df* plugin for
collectd. It should contain a map on key-value pairs that correspond
exactly to the plugin's configuration options.
*nvr0.pyrocufflink.blue* hosts Frigate. It is deployed on a separate
subnet, for two reasons:
* To avoid streaming video from the cameras through the firewall
* To prevent any hosts on the LAN except Home Assistant from
communicating with Frigate, since it does not have any kind of
authentication or access control
Frigate is an NVR that uses machine learning to detect objects on camera
in real time. It integrates with Home Assistant to expose sensors which
can be used for automation, etc.
The only official way to deploy Frigate is with a container, so we use
Podman and systemd to manage it.
For hosts that cannot send metrics via multicast (e.g. because they are
on a different subnet), *collectd* needs to listen on the all-hosts
unicast address.
The VM hosts have multiple network interfaces with IPv6 addresses, so
collectd may not always choose the correct one to send metrics. Thus we
have to explicitly tell it to use the management interface, to avoid it
sending data on the SAN interface.
The production deployment of *dnsmasq* for Home Assistant has deviated
from how the *hass-dhcp* role configures it. Bringing the role back in
sync with how things really are.
ZwaveJS2Mqtt includes a very powerful web-based UI for configuring and
controlling the Z-Wave network. This functionality is no longer
available within Home Assistant itself, so being able to access the
ZwaveJS2Mqtt UI is crucial to operating the network.
I wanted to make the UI available at */zwave/*, which requires using
*mod_rewrite* to conditionally proxy requests based on the `Connection`
HTTP header, since the UI passes both HTTP and WebSocket requests to the
same paths. *mod_rewrite* configuration is not inherited from the main
server configuration to virtual hosts, so the
`RewriteRule`/`RewriteCond` directives have to be specified within the
`<VirtualHost>` block. This means that the Home Assistant proxy
configuration has to be within its own virtual host, and the
Zwavejs2Mqtt configuration has to be there as well.
*hass2.pyrocufflink.blue* is a Raspberry Pi Compute Module 4-based
system, currently mounted in a WaveShare CM4 Mini Base Board (A). With
an NVMe SSD for primary storage, it runs significantly faster than a
standard Raspberry Pi 4, and blows the old Raspberry Pi 3-based Home
Assistant deployment out of the water. It has a Zooz 700 series Z-Wave
Plus S2 USB stick and a ConBee II Zigbee USB stick attached to its USB
2.0 ports. It runs a customized Fedora Minimal distribution.
Zigbee2MQTT is very similar to ZwaveJS2Mqtt: it is a daemon process that
communicates with the Zigbee radio and integrates with Home Assistant
using MQTT. Naturally, I decided to deploy it in the same way as
ZwaveJS2Mqtt, using a systemd unit to run it in a container with Podman.
Mosquitto 2.x included two significant changes from 1.6:
* There is no longer a "default" listener; all listeners are configured
in the same way
* The daemon drops privileges *before* reading TLS certificates and
private keys
Home Assistant no longer recommends using the built-in libopenzwave
integration for communicating with Z-Wave devices. Evidently, OpenZWave
is no longer maintained, and community efforts have shifted toward
Z-Wave JS.
Z-Wave JS is architecturally much different than the legacy Z-Wave
integration. Instead of running the network controller inside the Home
Assistant process, a separate daemon communicates with the Z-Wave radio.
Home Assistant integrates with that daemon using a WebSockets API. This
has the advantage of decoupling the network operation from the lifecycle
of the Home Assistant process: restarting Home Assistant (e.g. to load
new configuration changes) does not take the Z-Wave network offline.
ZwaveJS2Mqtt is a distribution of the Z-Wave JS daemon, as well as a
web-based user interface for configuring it. Although its name implies
that it uses MQTT for communication, this feature is actually optional,
and the native WebSockets API can still be used for integration with
Home Assistant.
I decided to follow the same deployment pattern for ZwaveJS2Mqtt as for
Home Assistant itself: run the application from a container image using
Podman. This of course simplifies the installation of the application
significantly, leaving most of that work up to the maintainer of the
container image. Podman provides the container runtime, managing the
privileges, etc. The systemd service unit starts Podman, configuring an
ephemeral container on each run. The container uses the default network
namespace, avoiding the unnecessary overhead of port mapping. It uses
Podman's "rootless" mode, via the `--uidmap` and `--gidmap` arguments,
mapping users inside the container, including root, to unprivileged
users on the host. The Z-Wave radio, which is specified by the
`zwavejs_device` Ansible variable, is passed into the container via the
`--device` argument.
Installing Home Assistant in a Python virtualenv is rather tedious,
especially on non-x86 machines. The main issue is Python packages that
include native extensions, as many of these do not have binary wheels
available for aarch64, etc. on PyPI. Thus, to install these, they have
to be built from source, which then requires the appropriate development
packages to be installed. Additionally, compiling native code on a
Raspberry Pi is excruciatingly slow. I have considered various ways of
mitigating this, but all would require a substantial time investment,
both up front and ongoing, making them rather pointless. Eventually, I
settled on just deploying the official Home Assistant container image
with Podman.
Although Podman includes a tool for generating systemd service unit
files for running containers, I ended up creating my own for several
reasons. First and foremost, the generated unit files configure the
containers to run as *root*, but I wanted to run Home Assistant as an
unprivileged user. Unfortunately, I could not seem to get the container
to work when dropping privileges using the `User` directive of the unit.
Fortunately, `podman` has `--uidmap` and `--gidmap` arguments, which I
was able to use to map UID/GID 0 in the container to the *homeassistant*
user on the host. Another drawback of the generated unit files is that
they specify a "forking" type service, which is not really necessary.
Podman/conmon supports the systemd notify protocol, but the generator
has not been updated to make use of that yet.
Recent versions of Home Assistant are more strict with respect to how
reverse proxies are handled. In order to use one, it must be explicitly
listed in the configuration file. Therefore, the *homeassistant*
Ansible role will now create a stub `configuration.yaml`, based on the
one generated by Home Assistant itslf when it starts for the first time
on a new machine, that includes the appropriate configuration for a
reverse proxy running on the same machine. The stub configuration will
not overwrite an existing configuration file, so it is only useful when
deploying Home Assistant for the first time on a new machine.
Overall, although I think a 300+ MB container image is ridiculous,
deploying Home Assistant this way should make it a lot easier to manage,
especially when updating.
Zezere is the Fedora IoT device provisioning service. It is the
software that runs *provision.fedoraproject.org*, but it can be
self-hosted (if you can figure it out; there is no documentation
whatsoever).
The main use case for running Zezere locally is to automatically add
trusted SSH public keys to Fedora IoT devices, without depending on a
cloud service. This playbook sets up Zezere with the very minimal
configuration needed to meet this goal.
Although configuration policy is not yet available for Prometheus
itself, the `collectd.yml` playbook also uses the *prometheus* host
group. Specifically, hosts in this group are configured to receive
collectd data from other hosts and expose those data through the
`write_prometheus` plugin.
This commit introduces the *grafana* role and the corresponding
`grafana.yml` playbook. The role installs Grafana using the system
package manager, and configures the server (including LDAP
authentication).
Occasionally, ProtonVPN servers randomly reject the EAP authentication
credentials. When this happens, the tunnel fails and is not restarted
automatically by strongSwan. As such, the watchdog needs to react to
this event as well.
Since the Nextcloud configuration file is managed by the configuration
policy, all of the settings configurable through the web UI need to be
templated. One important group of settings is the outbound email
configuration. This can now be configured using the `nextcloud_smtp`
Ansible variable.
This simple role installs the *redis* package and starts the associated
service. It leaves the configuration as provided by upstream, at least
for now.
Fedora now includes a packaged version of Nextcloud. This will be
_much_ easier to maintain than the tarball-based distribution method.
There are some minor differences in how the Fedora package works,
compared to the upstream tarball. Notably, it puts the configuration
file in `/etc/` and makes it read-only, and it stores persistent data
separate from the application. These differences require modifications
to the Apache and PHP-FPM configuration, but the package also included
examples to make this easier. Since the `config.php` is read-only now,
it has to be managed by the configuration policy; it cannot be modified
by the Administration web UI.
One major problem with the current DNS-over-VPN implementation is that
the ProtonVPN servers are prone to random outages. When the server
we're using goes down, there is not a straightforward way to switch to
another one. At first I tried creating a fake DNS zone with A records
for each ProtonVPN server, all for the same name. This ultimately did
not work, but I am not sure I understand why. strongSwan would
correctly resolve the name each time it tried to connect, and send IKE
initialization requests to a different address each time, but would
reject the responses from all except the first address it used. The
only way to get it working again was to restart the daemon.
Since strongSwan is apparently not going to be able to handle this kind
of fallback on its own, I decided to write a script to do it externally.
Enter `protonvpn-watchdog.py`. This script reads the syslog messages
from strongSwan (via the systemd journal, using `journalctl`'s JSON
output) and reacts when it receives the "giving up after X tries"
message. This message indicates that strongSwan has lost connection to
the current server and has not been able to reestablish it within the
retry period. When this happens, the script will consult the cached
list of ProtonVPN servers and find the next one available. It keeps
track of the ones that have failed in the past, and will not connect to
them again, so as not to simply bounce back-and-forth between two
(possibly dead) servers. Approximately every hour, it will attempt to
refresh the server list, to ensure that the most accurate server scores
and availability are known.
*Mosquitto* implements an MQTT server. It is the recommended
implementation for using MQTT with Home Assistant.
I have added this role to deploy Mosquitto on the Home Assistant server.
It will be used to send data from custom sensors, such as the
temperature/pressure/humidity sensor connected to the living room wall
display.
Since there are no other plain HTTP virtual hosts, the one defined for
chmod777.sh became the "default." Since it explicitly redirects all
requests to https://chmod777.sh, it caused all non-HTTPS requests to be
redirected there, regardless of the requested name. This was
particularly confusing for Tabitha, as she frequently forgets to put
https://…, and would find herself at my stupid blog instead of
Nextcloud.
When there is a network issue that prevents DNS names from being
resolved, it can be difficult to troubleshoot. For example, last night,
the Samba domain controller crashed, so *pyrocufflink.blue* names were
unavailable. Furthermore, the domain controller VM was apparently
locked up, so I could not SSH into it directly, and it needed to be
rebooted. Since the VM host's name did not resolve, I could not find
its address to log into it and reboot the VM. I resorted to scanning
the SSH keys of every IP address on the network until I found the one
that matched the cached key in ~/.ssh/known_hosts. This was cumbersome
and annoying.
Assigning DHCP reservations to the VM hosts will ensure that when a
situation like this arises again, I can quickly connect to the correct
VM host and manage its virtual machines, as its address is recorded in
the configuration policy.
The *esp4* kernel module does not load automatically on Fedora. Without
this module, strongSwan can establish IKE SAs, but not ESP SAs. Listing
the module name in a file in `/etc/modules-load.d` configures the
*systemd-modules-load* service to load it at boot.
Dustin