Fabric device_import playbook does not onboard interfaces when delete config is still removing config

[Sai Chakravarthy Alikapati]

The device_import playbook does not onboard interfaces of a device when the fabric delete is still removing config, but the discovery and role assignment still is successful. This happens when the discovery process is started as soon as the delete fabric job exits. The reason is that the playbooks are not able to edit the config on the device in private mode. It is the resultant of overlapping config edits on the devices. Following are the errors observed in the process.

fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg": "Unable to open the configuration in private mode: RpcError(severity: warning, bad_element: None, message: warning: changes cannot be committed while 'configure exclusive' is active\nwarning: uncommitted changes will be discarded on exit)"}

fatal: [localhost]: FAILED! => {"changed": false, "failed": true, "msg": "Unable to open the configuration in exclusive mode: LockError(severity: error, bad_element: None, message: configuration database modified)"}

Workaround for this issue is that the discovery process should not be started until the delete process is complete.

[Jeba Paulaiyan (jebap)]

Notes:

Between time intensive device manager operations (eg., device discovery, fabric role assignment, fabric delete, etc.) on the fabric, it is recommended to wait for sufficient time and make sure the previous operation is done.

[Akshaya (akshayam)]

In order to fix this issue, we can create a device level lock, based on device_fqname to indicate the device is locked and some operation is being performed on it. So, we can also test the lock to check if the device is free of any ongoing operations and then set the lock on that device before beginning the job. This change will be part of another big change. (This change would be included as part of the job manager in DM (migrating the job manager from fabric to DM)).