Smarter Device Manager¶
Manage hardware resource allocation without a need for privileged containers.
This chart is not maintained by the upstream project and any issues with the chart should be raised here
Source Code¶
Requirements¶
Kubernetes: >=1.22.0-0
Dependencies¶
| Repository | Name | Version |
|---|---|---|
| https://bjw-s.github.io/helm-charts | common | 1.5.1 |
Installing the Chart¶
To install the chart with the release name smarter-device-manager
OCI (Recommended)¶
Traditional¶
helm repo add gabe565 https://charts.gabe565.com
helm repo update
helm install smarter-device-manager gabe565/smarter-device-manager
Uninstalling the Chart¶
To uninstall the smarter-device-manager deployment
The command removes all the Kubernetes components associated with the chart including persistent volumes and deletes the release.
Configuration¶
Read through the values.yaml file. It has several commented out suggested values. Other values may be used from the values.yaml from the bjw-s common library.
Specify each parameter using the --set key=value[,key=value] argument to helm install.
helm install smarter-device-manager \
--set env.TZ="America/New York" \
gabe565/smarter-device-manager
Alternatively, a YAML file that specifies the values for the above parameters can be provided while installing the chart.
Custom configuration¶
Please consider overriding the default configuration through values.yaml. For example:
config: |
- devicematch: ^snd$
nummaxdevices: 20
- devicematch: ^rtc0$
nummaxdevices: 20
- devicematch: ^video[0-9]*$
nummaxdevices: 20
- devicematch: ^ttyACM[0-9]*$
nummaxdevices: 1
The above configuration allows up to twenty pods to make use of /dev/snd, /dev/rtc0 and each of /dev/video[0-9]* devices. Only one pod at a time can use any of the discovered host /dev/ttyACM[0-9]* devices. This means that if a pod has reserved a dongle at /dev/ttyACM0, another pod requesting the same device will stay pending.
Once the devices are discovered, they will be added as allocatable resources to the nodes:
$ kubectl describe nodes
...
Capacity:
cpu: 4
memory: 16104560Ki
pods: 110
smarter-devices/gpiochip0: 0
smarter-devices/i2c-0: 0
smarter-devices/snd: 20
smarter-devices/ttyUSB-Z-Stick-Gen5: 1
The hardware is requested by pods through resources, e.g.:
In this case host device /dev/ttyUSB-Conbee-2 will be given at the same path, and the pod will only be created when this device is available.
Values¶
Important: When deploying an application Helm chart you can add more values from the bjw-s common library chart here
| Key | Type | Default | Description |
|---|---|---|---|
| config | string | See values.yaml | Override default configuration [ref] |
| hostNetwork | bool | true | When using hostNetwork make sure you set dnsPolicy to ClusterFirstWithHostNet |
| image.pullPolicy | string | "IfNotPresent" | image pull policy |
| image.repository | string | "registry.gitlab.com/arm-research/smarter/smarter-device-manager" | image repository |
| image.tag | string | "v1.20.11" | image tag |
| persistence | object | See values.yaml | Configure persistence settings for the chart under this key. |
| priorityClassName | string | "system-node-critical" | Custom priority class for different treatment by the scheduler Setting this is not necessary, but it is recommended. [ref] |
| securityContext | object | See values.yaml | Configure the securityContext for this pod [ref] |
| termination.messagePath | string | "/var/log/termination-log" | Configure the path at which the file to which the main container's termination message will be written. Overrides the default of /dev/termination-log to allow read-only persistence.devfs at /dev. [ref] |
| termination.messagePolicy | string | "FallbackToLogsOnError" | Indicate how the main container's termination message should be populated. Valid options are File and FallbackToLogsOnError. smarter-device-manager does not support a termination-log, so use the container's log. [ref] |
Autogenerated from chart metadata using helm-docs