[ADAM GRIFFITH]

@Xusheng, the rough Python code is tailored to work with my IntelliJ IDE. However it does contain good examples of what environmental variables FABLib expects, how to call functions in FABLib, etc.. In short, I would write your own Python by grabbing the useful bits from mine.

[ADAM GRIFFITH]

My advice would be to start with copying the FABLib commands documented on the start_here.ipynb --> create_slice.ipynb Jupyter Lab notebook to your local Python script file.

If you would like, feel free to cannibalize my very rough Python code on GitHub.

[ADAM GRIFFITH]

If you are looking to interact with Fabric via Python I have had a lot of success with the FABLib API.

Here is our Conda environment.yml file:

# conda env create --file environment.yml
name: uofu-fabrictestbed
channels:
  - defaults
dependencies:
  - pip=21.2.4
  - python=3.9.12
  - pip:
      - fabrictestbed-extensions==1.3.0

[ADAM GRIFFITH]

While testing this in Jupyter I am yes. cat .tokens.json does show an up-to-date created_at but the oscillation between the group membership or invalid refresh token errors remain. Komal has been assigned my ticket and hopefully he can do some behind the scenes magic to clear this up.

[ADAM GRIFFITH]

@Paul I am following your Configure your Jupyter Environment docs.

I believe all the variables are present below.

cluster_logger.info('Setting up OS environmental variables for FABlib...')
os.environ['FABRIC_CREDMGR_HOST'] = 'cm.fabric-testbed.net'
os.environ['FABRIC_ORCHESTRATOR_HOST'] = 'orchestrator.fabric-testbed.net'
os.environ['FABRIC_BASTION_HOST'] = 'bastion-1.fabric-testbed.net'
os.environ['FABRIC_PROJECT_ID'] = config['secrets']['fabric_project_id']
os.environ['FABRIC_BASTION_USERNAME'] = config['secrets']['fabric_bastion_username']
os.environ['FABRIC_BASTION_KEY_LOCATION'] = f"{fabric_config_dir}/fabric-bastion-key"
os.environ['FABRIC_SLICE_PRIVATE_KEY_FILE'] = f"{fabric_config_dir}/slice_key"
os.environ['FABRIC_SLICE_PUBLIC_KEY_FILE'] = f"{fabric_config_dir}/slice_key.pub"
os.environ['FABRIC_LOG_LEVEL'] = 'INFO'
os.environ['FABRIC_LOG_FILE'] = f"{fabric_config_dir}/fablib.log"
os.environ['FABRIC_TOKEN_LOCATION'] = f"{fabric_config_dir}/token.json"

That membership error does shift to an “invalid refresh token” that won’t go away despite regenerating new token JSON files (submitted as FIP-91).

[ADAM GRIFFITH]

Is there a work-around for this error outside of Jupyter Lab? I download a fresh token from the Portal, set FABRIC_TOKEN_LOCATION and received:

Traceback (most recent call last):
  File "~/repos/fabric-testbed-api/cluster/main.py", line 75, in <module>
    raise ex
  File "~/repos/fabric-testbed-api/cluster/main.py", line 72, in <module>
    slice_id = fabric_slice.submit()
  File "~/.conda/envs/slate-fabrictestbed/lib/python3.9/site-packages/fabrictestbed_extensions/fablib/slice.py", line 1254, in submit
    return_status, slice_reservations = self.fablib_manager.get_slice_manager().create(slice_name=self.slice_name,
  File "~/.conda/envs/slate-fabrictestbed/lib/python3.9/site-packages/fabrictestbed/slice_manager/slice_manager.py", line 200, in create
    self.refresh_tokens()
  File "~/.conda/envs/slate-fabrictestbed/lib/python3.9/site-packages/fabrictestbed/slice_manager/slice_manager.py", line 160, in refresh_tokens
    raise SliceManagerException(tokens.get(CredmgrProxy.ERROR))
fabrictestbed.slice_manager.slice_manager.SliceManagerException: error: User is not a member of project: 'XXXX', refresh_token: abcdefg1234

[ADAM GRIFFITH]

Oops, missing fablib.show_config() for the first line above.

[ADAM GRIFFITH]

Hi Paul,

I was able to successfully lease a slice.

Thank you!
-Adam

[ADAM GRIFFITH]

Hi Mert,

Apologies. Here is the Python I can use to reproduce the stall at “Configuring”:

`
from fabrictestbed.slice_editor import ComponentModelType

slice_name = ‘KubernetesSlice13-adam’
site = ‘MAX’
node1_name = ‘Node1’
node2_name = ‘Node2′
network_service_name=’bridge1’
nic1_name = ‘node1-nic1’
nic2_name = ‘node2-nic1’
username = ‘centos’
image = ‘default_centos_8’
image_type = ‘qcow2′
cores = 4
ram = 20
disk = 100

nvme_name=’nvme1’
nvme_model_type = ComponentModelType.NVME_P4510
`

`
from fabrictestbed.slice_editor import ExperimentTopology, Capacities, ComponentType, ComponentModelType, ServiceType
# Create topology
t = ExperimentTopology()

# Add node
n1 = t.add_node(name=node1_name, site=site)

# Set capacities
cap = Capacities()
cap.set_fields(core=cores, ram=ram, disk=disk)

# Set Properties
n1.set_properties(capacities=cap, image_type=image_type, image_ref=image)

# Add node
n2 = t.add_node(name=node2_name, site=site)

# Set properties
n2.set_properties(capacities=cap, image_type=image_type, image_ref=image)

# Shared Cards
n1.add_component(model_type=ComponentModelType.SharedNIC_ConnectX_6, name=nic1_name)
n2.add_component(model_type=ComponentModelType.SharedNIC_ConnectX_6, name=nic2_name)

# Add the PCI NVMe device
n2.add_component(model_type=nvme_model_type, name=nvme_name)

# L2Bridge Service
t.add_network_service(name=network_service_name, nstype=ServiceType.L2Bridge, interfaces=t.interface_list)

# Generate Slice Graph
slice_graph = t.serialize()

# Request slice from Orchestrator
return_status, slice_reservations = slice_manager.create(slice_name=slice_name,
slice_graph=slice_graph,
ssh_key=ssh_key_pub)

if return_status == Status.OK:
slice_id = slice_reservations[0].get_slice_id()
print(“Submitted slice creation request. Slice ID: {}”.format(slice_id))
else:
print(f”Failure: {slice_reservations}”)
`

Cheers,
-Adam

[ADAM GRIFFITH]

Extended answer: For my personal sanity I added a pattern resembling to ~/.ssh/config file:

### The External Fabric Bastion host
Host fabric-bastion-host
  HostName bastion-1.fabric-testbed.net
  Port 22
  User susanna_moser_0051524387
  IdentityFile ~/work/id_rsa_fabric

This transforms ssh commands into ssh -J fabric-bastion-host -i ~/.ssh/id_rsa centos@63.239.135.94 without the need for the ssh-agent voodoo.

[ADAM GRIFFITH]

Try eval ssh-agent -s followed by ssh-add /path/to/<fabric key file>. Then take your ssh command above, remove the first key option/value and run it again.

• This reply was modified 2 years, 4 months ago by ADAM GRIFFITH.

[ADAM GRIFFITH]

Hi Paul,

Unfortunately it does not. I’ll ping Mina and see if he can make it through /jupyter-examples/fabric_examples/complex_recipes/kubernetes-simple.ipynb.

 

Cheers,
-Adam

[ADAM GRIFFITH]

Looks like things are back up.

[ADAM GRIFFITH]

This problem is present for me as well.

[Author]

Posts