cleanup the yaml file for worker in the k8s deployment

[smartfog]

In the yaml/cloud-yaml/worker.yaml and yaml/edge-yaml/edge-worker.yaml, the docker.sock part is not needed any more. They should be cleaned up.

• name: dockersock mountPath: "/var/run/docker.sock"

[showersky]

It seems that the usage of docker APIs is not complete cleaned up in the following file.

https://github.com/smartfog/fogflow/blob/development/k8sWorker/executor.go

This is why the k8s deployment script still includes the dockersock. We need to clean up the executor.go and replace all the usage of docker API with k8s API.

[BabutaAniket]

@showersky I have an doubt. Docker daemon is listening to docker.sock and it entry point for Docker API. In yaml/cloud-yaml/worker.yaml we mount Docker socket which enables us to launch new containers from within worker. So if I remove docker.sock from yaml files, my containers won’t launch while running any of the use-case:

root@ss:~/helm_fogflow# docker logs -f  fe920f3524f8
INFO: 2022/01/21 10:44:15 communicating with the broker  http://172.30.48.24:8070/ngsi10
INFO: 2022/01/21 10:44:15 annouce myself to the nearby broker
2022/01/21 10:44:15 [*] Waiting for messages. To exit press CTRL+C
INFO: 2022/01/21 10:47:08 I am going to fetch the docker image fogflow/nodejs
DEBUG: 2022/01/21 10:47:08 executor.go:117: options : {Username: Password: Email: ServerAddress: IdentityToken: RegistryToken:}
DEBUG: 2022/01/21 10:47:08 executor.go:124: options : {Repository:fogflow/nodejs Tag:latest Platform: Registry: OutputStream:<nil> RawJSONStream:false InactivityTimeout:0s Context:<nil>}
ERROR: 2022/01/21 10:47:08 executor.go:128: failed to pull this image fogflow/nodejs, error dial unix /var/run/docker.sock: connect: no such file or directory
INFO: 2022/01/21 10:50:35 execute task  Task.Crop_Prediction.Crop.1242463947  with operation fogflow/soil:latest
INFO: 2022/01/21 10:50:35 task configuration {ID:Task.Crop_Prediction.Crop.1242463947 ServiceName:Crop_Prediction TaskName:Crop OperatorName:Crop_Predict TaskType: FunctionCode: DockerImage:fogflow/soil:latest Parameters:[] WorkerID:Worker.001 IsExclusive:false PriorityLevel:50 Status:scheduled Inputs:[{Type:SoilSensor ID:urn:ngsi-ld:Device.SoilSensor54@default FiwareServicePath: MsgFormat:NGSILD AttributeList:[]}] Outputs:[{Type:CropPrediction StreamID:CropPrediction.1242463947.1 Annotations:[]}]}
INFO: 2022/01/21 10:50:35 task runnable =  true
&{Task.Crop_Prediction.Crop.1242463947 Crop_Prediction Crop Crop_Predict   fogflow/soil:latest [] Worker.001 false 50 running [{SoilSensor urn:ngsi-ld:Device.SoilSensor54@default  NGSILD []}] [{CropPrediction CropPrediction.1242463947.1 []}]}
INFO: 2022/01/21 10:50:35 to execute Task  Task.Crop_Prediction.Crop.1242463947  to perform Operation  fogflow/soil:latest
INFO: 2022/01/21 10:50:35 operator image fogflow/soil:latest does not exist locally
DEBUG: 2022/01/21 10:50:35 executor.go:117: options : {Username: Password: Email: ServerAddress: IdentityToken: RegistryToken:}
DEBUG: 2022/01/21 10:50:35 executor.go:124: options : {Repository:fogflow/soil:latest Tag:latest Platform: Registry: OutputStream:<nil> RawJSONStream:false InactivityTimeout:0s Context:<nil>}
**ERROR: 2022/01/21 10:50:35 executor.go:128: failed to pull this image fogflow/soil:latest, error dial unix /var/run/docker.sock: connect: no such file or directory**
ERROR: 2022/01/21 10:50:35 executor.go:311: failed to fetch the image fogflow/soil:latest

I have tried changing the endpoint to other CRI in k8s but the issue is still present. In /k8sworker/pod.go file, the image for the in pod is passed from executor.go and in executor.go the endpoint for unix is set default to unix:///var/run/docker.sock. Even if I replace docker API with k8s API in executor.go, it is difficult to launch a container inside another container without docker.sock (same error as above).

[showersky]

this has been addressed in the latest deployment script for k8s