Add error handling and tests

[dschniepp]

Over christmas we experienced (CaseClauseError) no case clause matching: {:error, :econnrefused} errors.

Elixir.CaseClauseError: no case clause matching: {:error, :econnrefused}
  File "lib/libcluster_ec2.ex", line 53, in ClusterEC2.request/1
  File "lib/libcluster_ec2.ex", line 29, in ClusterEC2.local_instance_tags/0
  File "lib/libcluster_ec2.ex", line 38, in ClusterEC2.local_instance_tag_value/1
  File "lib/strategy/tags.ex", line 96, in ClusterEC2.Strategy.Tags.get_nodes/1
  File "lib/strategy/tags.ex", line 61, in ClusterEC2.Strategy.Tags.handle_info/2
...
(4 additional frame(s) were not displayed)

(CaseClauseError) no case clause matching: {:error, :econnrefused}

So I changed the code in order to prevent exceptions. What do you think?

• Reimplemented request via Tesla
• Added tests
• Moved tags related code from lib/libcluster_ec2.ex to lib/strategy/tags.ex
• added .editorconfig

[kyleaa]

This looks great!

I'm curious and haven't found it documented - do you know if returning an empty list from get_nodes will cause all of the nodes to disconnect?

[dschniepp]

Good point, I think it will cause that all nodes get disconnected otherwise Cluster.Strategy.disconnect_nodes wouldn't be necessary.

Thank you for pointing out I will propose a solution. Maybe we can bubble up the errors and in case of a connection error or timeout we will keep the state as it is.

[dschniepp]

@kyleaa i changed the logic to the following behaviour:
In case the instance_id or region is not set the node_list will not be fetched and returns an error tuple. If the error appears the existing state won't be changed. What do you think?