Fix nxos and nxos_ssh hostname change handling

[ktbyers]

I ran into issues with both nxos driver and nxos_ssh driver with hostname changes. Made fixes here to incorporate.

I also updated the napalm_custom_test suite to test this for both nxos and nxos_ssh for both merge and replace operations. This generally runs daily against virtual network devices and the develop branch.

[coveralls]

Coverage decreased (-0.01%) to 80.658% when pulling 8c6c833 on ktbyers:nxos_hostname into 7980208 on napalm-automation:develop.

[coveralls]

Coverage decreased (-0.01%) to 80.658% when pulling 8c6c833 on ktbyers:nxos_hostname into 7980208 on napalm-automation:develop.

[coveralls]

Coverage decreased (-0.01%) to 80.658% when pulling 8c6c833 on ktbyers:nxos_hostname into 7980208 on napalm-automation:develop.

[lazhao123]

The bug has not fixed yet. I changed hostname with load_merge_candidate( ), the base_prompt is not changed, so check_enable_mode( ) still search old hostname and timeout. Attached stack trace.

        if self.replace:
            self._load_cfg_from_checkpoint()
            # If hostname changes ensure Netmiko state is updated properly
            self._netmiko_device.set_base_prompt()         <========= should call when merge config
        else:
            self._commit_merge()

Error thrown from base_connection.py: "Timed-out reading channel, data not available."
except socket.timeout:
raise NetMikoTimeoutException(
"Timed-out reading channel, data not available."
)

[lazhao123]

move set_base_prompt( ) before self._copy_run_start( ) fixed it.

    if self.loaded:
        # Create checkpoint from current running-config
        self._save_to_checkpoint(self.rollback_cfg)

        if self.replace:
            self._load_cfg_from_checkpoint()
        else:
            self._commit_merge()

        # If hostname changes ensure Netmiko state is updated properly
        self._netmiko_device.set_base_prompt()
        self._copy_run_start()

[lazhao123]

After rollback( ), prompt gets changed, but "copy_run_start( )" still look for old prompt, that caused timeout.
So I added set_base_promt( ) after rollback to fixed it.

def rollback(self):
    if self.changed:
        command = "rollback running-config file {}".format(self.rollback_cfg)
        result = self._send_command(command)
        if "completed" not in result.lower():
            raise ReplaceConfigException(result)
        # If hostname changes ensure Netmiko state is updated properly   
        self._netmiko_device.set_base_prompt()             <=== change prompt before "copy run st"
        self._copy_run_start()
        self.changed = False

[ktbyers]

@lazhao123 Okay, I opened the referenced new issue.

Yes, this is still broke. My test case was working for merge, but it was because I chose poorly on the new hostname (i.e. it had part of the old prompt in the new prompt name).

There is a good chance your fix for the merge part of this won't work (i.e. I think that is probably going to break NX-API as there is no Netmiko connection there for a merge operation...need to double check this though).