fog list fails if instance is not fully set up yet

[mjd3]

Describe the bug
If the instance is still starting up, running ros2 fog list results in a FileNotFoundError.

To Reproduce
Steps to reproduce the behavior:

1. Run ros2 launch fogros2_examples talker.launch.py
2. While this is starting the EC2 instance, run ros2 fog list
3. Errors with stack trace:

Expected behavior
Should output a list of current fogros2 instances.

[KeplerC]

The current design is that we only flush the info file after the instance is launched, but we still create a directory for storing temporary files such as pem keys.

Is there any better way of doing this? (I know I need to add error handling)

[mjd3]

Yeah I think there are multiple options here: 1) just check if the info file is not available yet and return an abbreviated description for this instance, 2) remove any instances from ros2 fog list that are not completely set up yet. I'd probably lean towards (1) but as long as it doesn't crash it's probably fine.

[jeffi]

The fix for #33 now reuses a single security group for all runs. We can use it to filter EC2 instances with the with fog list and related CLI commands.

  instances = boto3_client.describe_instances(
    Filters=[{"Name":"instance.group-name", "Values":["FOGROS2_SECURITY_GROUP"]}])

We could also consider using other AWS/EC2 functionality like tags.

With this, we can get most of the information related to an instance other than the private key. So we still have to store something locally. Also need to consider how we can do something similar with other providers (e.g., GCP).

[KeplerC]

the check existence is resolved by 7807f26. We should close this issue

[jeffi]

I'm working on a more robust version that queries AWS directly.

[jeffi]

Updated CLI to use AWS API directly. Added some modifications to the launch process to support this change. FogROS 2 instances get names similar to Docker instances. EC2 instances now have a FogROS2-Name tag with their name. Associated commits: d15bb33 and d4f70e8.