Open wang542 opened 5 years ago
great. The azure project is ongoing and a student is still working on it. see the checkins in the cloudmesh-cloud repo on azure. Missing features I am aware of are
The first one is very well documented and pointe out to the student.
We have a slack channel at https://app.slack.com/client/TFQC2P62D/CJWRBF1D3/thread/CFQSP8Q4T-1567738729.016200 in which that particular student reports progress. Maybe you guys can contact him his name is Joaquin
The Azure Provider class has implemented all the methods required by the ComputeNodeABC Abstract Class.
There are 3 methods related to key management where I got stuck:
def keys(self):
raise NotImplementedError
def key_upload(self, key=None):
raise NotImplementedError
def key_delete(self, name=None):
raise NotImplementedError
The main method is key_upload which takes the key from your Operating System and uploads it to the cloudmesh selected cloud.
Other clouds like openstack or aws have a pretty straightforward implementation that requires two parameters (key_name and public_key)
I.E:
OPENSTACK
self.cloudman.create_keypair(name, key['public_key'])
AWS
self.ec2_client.import_key_pair(KeyName=key_name,PublicKeyMaterial=key['public_key'])
Unfortunately I did not find anything similar on the Azure SDK. Azure uses a template that defines each Virtual Machine set of Parameters.
I started with a structure like the one below.
vm_parameters = {
'location': self.LOCATION,
'os_profile': {
'computer_name': self.VM_NAME,
'admin_username': self.USERNAME,
'admin_password': self.PASSWORD
},
'hardware_profile': {
'vm_size': self.VM_SIZE
},
'storage_profile': {
'image_reference': {
'publisher': publisher,
'offer': offer,
'sku': sku,
'version': version
},
},
'network_profile': {
'network_interfaces': [{
'id': nic.id,
}]
},
}
When looking for a way to upload the public key I found that it is supported under the os_profile tag's linux configuration.
FOR EXAMPLE
'os_profile': {
'computer_name': self.VM_NAME,
'admin_username': self.USERNAME,
"linux_configuration": {
"disable_password_authentication": True,
"ssh": {
"public_keys": [{
"path": "/home/{}/.ssh/authorized_keys".format(USERNAME),
"key_data": "ssh-rsa ABCD....."
}]
}
}
}
The following classes are used to assemble a full setup of vm_parameters:
VirtualMachine: -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.virtualmachine?view=azure-python this is the main main class that describes the Virtual Machine through the following parameters: Location Os Profile:-> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.osprofile?view=azure-python ... this class takes an attribute called Linux configuration... Linux Configuration -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.linuxconfiguration?view=azure-python ... then the Linux configuration takes an attribute called ssh configuration... Ssh Configuration -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.sshconfiguration?view=azure-python ... which ultimately uses another one called ssh public key which is the one that takes the key path and key data like other clouds... SshPublicKey -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.sshpublickey?view=azure-python HardwareProfile: -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.hardwareprofile?view=azure-python StorageProfile: -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.storageprofile?view=azure-python ImageReference: -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.imagereference?view=azure-python NetworkProfile: -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.networkprofile?view=azure-python NetworkInterfaceReference: -> https://docs.microsoft.com/en-us/python/api/azure-mgmt-compute/azure.mgmt.compute.v2019_03_01.models.networkinterfacereference?view=azure-python
The SshPublicKey class seemed to me like the equivalent implementation to AWS and Openstack.
It takes the following Parameters: path: Specifies the full path on the created VM where ssh public key is stored. If the file already exists, the specified key is appended to the file. Example: /home/user/.ssh/authorized_keys
key_data: SSH public key certificate used to authenticate with the VM through ssh. The key needs to be at least 2048-bit and in ssh-rsa format.
The difference is that it needs to be included as part of the virtual machine parameters setup.
I tried to assemble it using the following code but have not commited it since I have not been able to make it work.
from azure.mgmt.compute.v2019_03_01.models import OSProfile, LinuxConfiguration, SshConfiguration, SshPublicKey, VirtualMachine, HardwareProfile, StorageProfile, NetworkProfile, ImageReference, NetworkInterfaceReference
def create_vm_parameters(self):
nic = self.create_nic()
# Parse Image from yaml file
publisher, offer, sku, version = self.default["image"].split(":")
# Declare Virtual Machine Settings
ssh_keys = self.key_upload(self)
ssh_conf = SshConfiguration(public_keys=ssh_keys)
linux_conf = LinuxConfiguration(disable_password_authentication= True, ssh=ssh_conf, provision_vm_agent=False)
operating_system_profile = OSProfile(computer_name=self.VM_NAME, admin_username=self.USERNAME ,admin_password=self.PASSWORD,linux_configuration=linux_conf )
hardware_prof = HardwareProfile(vm_size=self.VM_SIZE)
image_ref = ImageReference( publisher = publisher, offer = offer, sku = sku, version= version)
storage_prof = StorageProfile(image_reference=image_ref)
network_int = NetworkInterfaceReference(id = nic.id)
network_prof = NetworkProfile(network_interfaces=network_int)
vm_parameters = VirtualMachine(
location=self.LOCATION,
os_profile = operating_system_profile,
hardware_profile = hardware_prof,
storage_profile = storage_prof,
network_profile = network_prof
)
return vm_parameters
Regarding project selection of Yanting hid: fa19-516-170 and myself hid: fa19-516-157. As you may recall we both wanted to do the cloudmesh-ai with Qiwei hid: fa19-516-151. It is my understanding that we would need azure compute service as a base of the cloudmesh-ai service, for comparing purposes. And we need someone to work on azure compute, otherwise cloudmesh-ai project can't take place.
After our discussion, I have decided to replace yanting in the azure compute project since yanting prefers to work on the cloudmesh-ai service. Since azure compute project is partially finished, naturally, I can continue to work on cloudmesh-ai after completion of azure compute or do some comparison of azure compute and other cloud providers.
Yanting and Qiwei both have been notified of this issue post. They will comment if necessary.