Configuration Notes
Specifications
Item |
Description |
|---|---|
Container |
By default, the AtlasEdge supports 16 containers. A maximum of 128 containers can be deployed through parameter configuration. You are not advised to modify the configuration because too many containers may deteriorate device performance. |
The available memory of a single container is 2048 MB by default. You can configure the memory specifications of a single container by setting parameters. Formula to calculate the total available memory resources of all containers: Total available memory resources = Total system memory resources – Reserved system memory resources |
|
By default, two CPU cores are available for a single container. You can configure the CPU specifications of a single container by setting parameters. Formula to calculate the total available CPU resources of all containers: Total available CPU resources = Total CPU resources of the system – Reserved CPU resources of the system |
|
By default, an NPU core is available for a single container. You can configure the NPU specifications of a single container by setting parameters. In virtual NPU scenarios, a single container supports a maximum of one NPU core. |
|
ConfigMap |
The number of ConfigMaps is four times the number of containers. That is, 64 ConfigMaps can be configured by default, and a maximum of 512 ConfigMaps can be configured. |
Docker daemon |
A single process can hold a maximum of 64 file handles at the same time. |
A single UID user supports a maximum of 512 fork processes. If multiple containers are started using the same UID, the maximum number of fork processes of multiple containers is 512. (16 containers use the same UID, and each container has 32 processes on average.) |
- Do not deploy containers with the same ID repeatedly. Otherwise, the container name may be incorrectly displayed.
- Restarting the AtlasEdge will change the creation time of the pod where the container is deployed on FusionDirector to the current time.
- Containers deployed using an NMS can be deleted only by the NMS. Do not delete the containers by running commands. Otherwise, services may be abnormal.
- After restarting the Docker service, restart the AtlasEdge.
- The AtlasEdge is incompatible with the kubelet and kube-proxy processes. If the kubelet or kube-proxy software exists in the environment and is running, the AtlasEdge components fail to be started. Therefore, check whether the kubelet and kube-proxy processes are installed in the environment. If they are installed, stop or uninstall them in advance. You can also install the AtlasEdge software on other devices without the kubelet and kube-proxy processes.