Security Group VM members unable to sync with OpenStack

[pravinrajr9]

Description

Security Group VM Member is not able to sync with OpenStack.
When a Security group is created with one or more VMs and when the VM is deleted through OpenStack Horizon, this should decrement the No of VM members count in Security group. The sync functionality passes and also fails. This Happens only with OpenStack Pike Version.

Expected Behavior

The Security Group VM member should be deleted.
In the Task SecurityGroupCheckMetaTask, SecurityGroupMemberDeleteTask function should be called

Actual Behavior

The Security Group VM member port is deleted, VM member is not deleted.
In the Task SecurityGroupCheckMetaTask, SecurityGroupMemberVmUpdateTask function is called.

Steps to Reproduce

1. Create VC with NSC as plugin
2. Create SG with any no of VMs
3. Go to OpenStack Horizon and delete the same VM
4. Check the Security group VM member count
   To reproduce the bug, create SG with more than one VM, and repeat step 3 multiple times until the bug is found

Additional Information

Workaround manually click sync button in Security group panel, the VM member count is updated.

Environment

• OSC Version/Build: Version: 0.8.0 (Build:1-g62f2b90, 2017-12-01T19:51:38Z)
• Cloud Provider Info:
• Plugin Info: NSC SDN Plugin
• Other: OpenStack Version: OpenStack Pike

Status

• Reproduced
• In Progress
• Validated => Verified in Version/Build:

ServerSupportBundle (10).zip

[sudhirappaji]

@bmmukund , please investigate and if needed, discussed with Dmitry.

[bmmukund]

I am looking into this issue.

[dmitryintel]

A sure way to reproduce is to create 4 or 6 servers, add them to the security group, and then delete them from command line on openstack in quick succession:

(openstack) server delete vm1
(openstack) server delete vm2
(openstack) server delete vm3
(openstack) server delete vm4

[dmitryintel]

Per discussion with the team, in order to fix this bug, we need to do more granular locking of objects on our sync jobs. E.g., in this case, we are deleting a single VM member. We should look into locking that instead of the entire security group.

This needs to be addressed at a more global scale, rather than fixing this particular bug. Closing the Pull Request for now. It is to be addressed in future releases, as a part of a larger story.