Ah, got it - the secret is created by the chart itself.
I am currently migrating from bash script orchestration to a Kubernetes cluster. I am using https://github.com/rwlove/billimek-charts/commits/master/charts/home-assistant (fork’d repo, see top 4 patches) to deploy my Home Assistant instance. I have been hacking away and I’ve created a patchset that I think adds useful functionality. I don’t think the series is ready to be merged and I’m creating this post to solicit feedback on my changeset.
New Functionality
- Ability to clone upstream custom components such that they’re loaded when Home Assistant starts.
- Clone upstream secrets.yaml file such that it’s loaded when Home Assistant starts.
- Mount /dev/wyze in the Pod for the Wyzesense Hub (similar to zwave)
- Improve git-sync initContainer to pull from git if the git repo already exists when cloning the Home Assistant configuration
Changes #1 and #2)
These patches presume you are already enabling git-sync to clone your configuration from an upstream git repository. They require the user to create the following symlinks in their configuration git repository that git-sync will clone or pull from. My current layout looks like this:
➜ homeassistant-configuration git:(master) ls -l
total 16
-rw-rw-r-- 1 rwlove rwlove 3 Jun 27 07:57 automations.yaml
-rw-rw-r-- 1 rwlove rwlove 613 Jun 27 07:57 configuration.yaml
lrwxrwxrwx 1 rwlove rwlove 20 Jun 27 07:57 custom_components -> ../custom_components
-rw-rw-r-- 1 rwlove rwlove 0 Jun 27 07:57 groups.yaml
-rw-rw-r-- 1 rwlove rwlove 29 Jun 14 12:05 README.md
-rw-rw-r-- 1 rwlove rwlove 0 Jun 27 07:57 scenes.yaml
-rw-rw-r-- 1 rwlove rwlove 0 Jun 27 07:57 scripts.yaml
lrwxrwxrwx 1 rwlove rwlove 23 Jun 28 17:37 secrets.yaml -> ../secrets/secrets.yaml
-rw-rw-r-- 1 rwlove rwlove 179 Jun 14 12:46 secrets.yaml.example
The patchset will create the /custom_components/ and /secrets/ directories. The user is also required to have a secrets.yaml file in their Home Assistant secrets upstream git repository.
Within the pod it creates the structure:
bash-5.0# ls /config/ -l
total 65688
-rw-r--r-- 1 root root 29 Jun 30 19:32 README.md
drwxr-xr-x 5 root root 94 Jun 30 19:34 appdaemon
-rw-r--r-- 1 root root 3 Jun 30 19:32 automations.yaml
-rw-r--r-- 1 root root 613 Jun 30 19:32 configuration.yaml
lrwxrwxrwx 1 root root 20 Jun 30 19:32 custom_components -> ../custom_components
drwxr-xr-x 2 root root 10 Jun 30 19:34 deps
-rw-r--r-- 1 root root 0 Jun 30 19:32 groups.yaml
-rw-r--r-- 1 root root 290 Jun 30 19:35 home-assistant.log
-rw-r--r-- 1 root root 139264 Jun 30 19:35 home-assistant_v2.db
-rw-r--r-- 1 root root 0 Jun 30 19:32 scenes.yaml
-rw-r--r-- 1 root root 0 Jun 30 19:32 scripts.yaml
lrwxrwxrwx 1 root root 23 Jun 30 19:32 secrets.yaml -> ../secrets/secrets.yaml
-rw-r--r-- 1 root root 179 Jun 30 19:32 secrets.yaml.example
drwxr-xr-x 2 root root 10 Jun 30 19:32 tts
My values.yaml looks like this for the above example:
secret:
enabled: true
## we just use the hass-configurator container image
## you can use any image which has git and openssh installed
##
image:
repository: causticlab/hass-configurator-docker
tag: 0.3.5-x86_64
pullPolicy: IfNotPresent
secret: git-creds
syncPath: /config
keyPath: /root/.ssh
repo: "[email protected]:rwlove/<my-private-secrets-repo>.git"
syncPath: /secrets
customcomponents:
enabled: true
## we just use the hass-configurator container image
## you can use any image which has git and openssh installed
##
image:
repository: causticlab/hass-configurator-docker
tag: 0.3.5-x86_64
pullPolicy: IfNotPresent
## Specify the command that runs in the git-sync container to pull in configuration.
# command: []
syncPath: /custom_components
components:
- name: ewelink
repo: "https://github.com/peterbuga/HASS-sonoff-ewelink"
componentPath: HASS-sonoff-ewelink/sonoff
- name: wyzesense
repo: "https://github.com/kevinvincent/ha-wyzesense"
componentPath: ha-wyzesense/custom_components/wyzesense
Change #3)
I currently use the below rule to create /dev/wyze. I also manually tag the node with the wyze hub USB stick.
➜ ~ cat /etc/udev/rules.d/99-wyze.rules
SUBSYSTEM=="usb", ATTRS{idVendor}=="1a86", ATTRS{idProduct}=="e024", SYMLINK+="wyze"
kubectl label nodes <node name> device=wyze
The relevant values.yaml:
wyze:
enabled: true
device: wyze
I can provide my full Home Assistant creation and deletion scripts, if anyone is curious.
Change #4)
This change is fairly simple. I use git to check if /config is a valid git repo. If it is, then I do a ‘git pull’, if not, I do a ‘git clone’.
Questions / Opens)
- I probably just want to pass in a command for the custom components. I’m doing a weird ‘mv’ in the initContianer because each custom components directory layout is different, but I’ve put too much logic in the initContainer. Simply providing a command in values.yaml for each custom component is probably more of a robust solution.
- I cannot seem to get HACS to install without a lot of manual changes. I have appdaemon running, but beyond that it’s unclear to me what I need to do. I can add it through my newly created custom components feature, but I have the impression that the functionality should already exist with appdaemon.
- Is there a better way to converge the multiple git repos being pulled into the Pod? I’m currently requiring the user to create symlinks in their configuration git repo, but maybe there is a better way.
TODO)
- Use .Values.git.rep and .Values.git.tag for customcomponents and secrets.
- Remove .Values.customcomponents.componentPath and just provide a .Values.customcomponents.command for each custom component.
- Update README.md
3 Likes
How are you guys getting sonos working inside your pods? Home assistant won’t discover sonos now that it’s in a different subnet, with a cluserip and load balanced ip
I guess you are looking for something like avahi-reflector or bonjour-reflector ?
Lars
Does anybody has values.yaml to deploy pvc on nfs ?
I try to understand how to deploy this
For those who are interested in running in K8s without privileged, and still use USB sticks, this worked for me (with custom naming through udev to /dev root): https://gitlab.com/arm-research/smarter/smarter-device-manager
This way you start a dedicated privileged container, which allows non-privileged pods to use selected devices, and manage their allocation.
I would love to kill hostnetwork and keep LIFX and uPnP, but the only way I found is through multus, which is in the edge of microk8s.
Any alternatives are welcome.
I also use Multus in my microk8s cluster to access my Google Home devices. I use the macvlan mode so I do not have to setup manually any bridge on the host.
I run on glusterfs and use an external postgres cluster (Zalando operator) so I can failover HA between cluster nodes.
It seems the helm chart was modified / moved ? I cannot get the code server sidecar to run - any hints on this ?
Thanks, Lars
@angelnu, what’s you overhead with running external postgress cluster? i.e. service memory and cpu utilisation? Is it used only for HA or anything else?
Hi Ingvarr,
Can you please share how are you able to use smarter-device-manager for bluetooth ?
Or maybe anybody has at least guide how to directly map it in k8s?
Need for bluetooth_le_module
Thank you,
I run the postgres in HA mode using the Zalando operator and a wrapping chart I wrote.
Currently the primary instance is consuming 0.7 CPUs and 200 MB RAM while the backup is 0.03 CPUs and 36 MB RAM. I do not recall how much I was consuming before with the local mysql.
As long as device is present as a file in /dev/ (root!), it can be just added to the configuration. In this helm chart, it is possible through values.
+count me in.
I’m currently working with Kubernetes and learning more everyday. The learning curve is steep.
I had HA running on RaspPi a couple years ago but got very frustrated that code upgrades would corrupt my systems and abandoned HA as a result.
From what I have read it is much more stable and given it will work on Kubernetes I am game to jump back into the pool…
My desire is to use ceph for persistent storage. I have a specific interest in this scenario.
Has anyone created this sort of setup?
One additional question is where will I find all the current helm charts and the kubernetes install instrutions? Several of the above links are old and no longer functioning.
Thanks.