Cancelled Scale workflow drops "started" nodes.


Steps to reproduce:

1. Install Cloudify Manager 4.3 with default and external network. "default" maps to same value as "private_ip", "external" maps to same value as "public_ip".

For example, from /etc/cloudify/config.yaml:
69 networks: {default:, external:}

2. Add your AWS credentials.

3. Add the utilities plugin and awssdk plugin .

4. Install the aws-network-example blueprint.

5. Install the aws nodecellar example, this version for 4.3.

6. Check out the node-instance list for the deployment. There are two `nodejs_host` nodes.

7. Execute the scale workflow on the `nodejs_group`. (It may hang, which is a different issue with this blueprint, mainly that the agent isn't communicating anymore with the manager. This may be another issue with 4.3, or it may be in this blueprint. I am not sure. In any case, this problem led me to discover the issue that I am currently describing.)

8. During the execution after a new nodejs_host node has been created, check the node instance list. You will see three nodejs_host node instances.

9. After the worfklow as beyond starting of the nodejs_host nodek, cancel the workflow.

_An unrelated issue: if you get to this step and cancel, the uninstall workflow will be stuck in pending for some reason:

10. Check the node-instances of nodejs_host. The new one from the scale execution is missing! Mysterious.

Expected behavior is that all existing nodes will remain in the model, so that they can be uninstalled.




Trammell -



Bug Type


Target Version




Affects versions