README v1.1.13 2025-09-26
Table of contents
1. General
1.1 Extract the NED package
1.2 Install the NED package
1.2.1 Local install
1.2.2 System install
1.3 Configure the NED in NSO
2. Optional debug and trace setup
3. Dependencies
4. Sample device configuration
5. Built in live-status actions
6. Built in live-status show
7. Limitations
8. How to report NED issues and feature requests
9. How to rebuild a NED
10. Configure the NED to use ssh multi factor authentication
11. Mandatory commands that have to be provided by the user
12. "security" section commands
13. How to configure additional config warning exceptions1. General
This document describes the cisco-fxos NED.
The cisco-fxos NED is built for 41xx/93xx devices. Also, the NED supports partial configuration for CISCO FXOS 21xx device.
The NED connects to the device CLI using SSH. The configuration is done by sending native CLI commands to the device through the communication channel.
FXOS uses a managed object model, where managed objects are abstract representations of physical or logical entities that can be managed. There are 4 commands available for object management:
create object
delete object
enter object
scope object
All these 4 keywords (create, delete, enter and scope) are internally handled by the NED, so they should NOT be mentioned by the user. Also, the keyword 'set' used for setting property values, is controlled by the NED and the user must skip it.
Additional README files bundled with this NED package
Common NED Features
Verified target systems
1.1 Extract the NED package
It is assumed the NED package ncs-<NSO version>-cisco-fxos-<NED version>.signed.bin has already been downloaded from software.cisco.com.
In this instruction the following example settings will be used:
NSO version: 6.0
NED version: 1.0.1
NED package downloaded to: /tmp/ned-package-store
Extract the NED package and verify its signature:
In case the signature can not be verified (for instance if no internet connection), do as below instead:
The result of the extraction shall be a tar.gz file with the same name as the .bin file:
1.2 Install the NED package
There are two alternative ways to install this NED package. Which one to use depends on how NSO itself is setup.
In the instructions below the following example settings will be used:
NSO version: 6.0
NED version: 1.0.1
NED download directory: /tmp/ned-package-store
NSO run time directory: ~/nso-lab-rundir
A prerequisite is to set the environment variable NSO_RUNDIR to point at the NSO run time directory:
1.2.1 Local install
This section describes how to install a NED package on a locally installed NSO (see "NSO Local Install" in the NSO Installation guide).
It is assumed the NED package has been been unpacked to a tar.gz file as described in 1.1.
Untar the tar.gz file. This creates a new sub-directory named:
cisco-fxos-<NED major digit>.<NED minor digit>:Install the NED into NSO, using the ncs-setup tool:
Open a NSO CLI session and load the new NED package like below:
Alternatively the tar.gz file can be installed directly into NSO. Then skip steps 1 and 2 and do like below instead:
Set the environment variable NED_ROOT_DIR to point at the NSO NED package:
1.2.2 System install
This section describes how to install a NED package on a system installed NSO (see "NSO System Install" in the NSO Installation Guide).
It is assumed the NED package has been been unpacked to a tar.gz file as described in 1.1.
Do a NSO backup before installing the new NED package:
Start a NSO CLI session and fetch the NED package:
Install the NED package (add the argument replace-existing if a previous version has been loaded):
Load the NED package
1.3 Configure the NED in NSO
This section describes the steps for configuring a device instance using the newly installed NED package.
Start a NSO CLI session:
Enter configuration mode:
Configure a new authentication group (my-group) to be used for this device:
Configure a new device instance (example: dev-1):
If configured protocol is ssh, do fetch the host keys now:
Finally commit the configuration
Verify configuration, using a sync-from.
If the sync-from was not successful, check the NED configuration again.
2. Optional debug and trace setup
It is often desirable to see details from when and how the NED interacts with the device(Example: troubleshooting)
This can be achieved by configuring NSO to generate a trace file for the NED. A trace file contains information about all interactions with the device. Messages sent and received as well as debug printouts, depending on the log level configured.
NSO creates one separate trace file for each device instance with tracing enabled. Stored in the following location:
$NSO_RUNDIR/logs/ned-cisco-fxos-cli-1.0-<device name>.trace
Do as follows to enable tracing in one specific device instance in NSO:
Start a NSO CLI session:
Enter configuration mode:
Enable trace raw:
Alternatively, tracing can be enabled globally affecting all configured device instances:
Configure the log level for printouts to the trace file:
Alternatively the log level can be set globally affecting all configured device instances using this NED package.
The log level 'info' is used by default and the 'debug' level is the most verbose.
IMPORTANT: Tracing shall be used with caution. This feature does increase the number of IPC messages sent between the NED and NSO. In some cases this can affect the performance in NSO. Hence, tracing should normally be disabled in production systems.
An alternative method for generating printouts from the NED is to enable the Java logging mechanism. This makes the NED print log messages to common NSO Java log file.
$NSO_RUNDIR/logs/ncs-java-vm.log
Do as follows to enable Java logging in the NED
Start a NSO CLI session:
Enter configuration mode:
Enable Java logging with level all from the NED package:
Configure the NED to log to the Java logger
Alternatively Java logging can be enabled globally affecting all configured device instances using this NED package.
IMPORTANT: Java logging does not use any IPC messages sent to NSO. Consequently, NSO performance is not affected. However, all log printouts from all log enabled devices are saved in one single file. This means that the usability is limited. Typically single device use cases etc.
SSHJ DEBUG LOGGING For issues related to the ssh connection it is often useful to enable full logging in the SSHJ ssh client. This will make SSHJ print additional log entries in $NSO_RUNDIR/logs/ncs-java-vm.log:
3. Dependencies
This NED has the following host environment dependencies:
Java 1.8 (NSO version < 6.2)
Java 17 (NSO version >= 6.2)
Gnu Sed
Dependencies for NED recompile:
Apache Ant
Bash
Gnu Sort
Gnu awk
Grep
Python3 (with packages: re, sys, getopt, subprocess, argparse, os, glob)
4. Sample device configuration
How to set a 'pre-login-banner' message
When configuring a 'pre-login-banner' you must assure that the line ends with a '\r\n'. In other words, the minimum size for a banner line is: '\r\n'. Using the 'commit dry-run outformat-native' command, will display how the command that will be send to the device will look like.
'set message' and 'ENDOFBUF' commands are automatically added.
Notes:
On the device the following 2 commands are equivalent: clear message and set message >ENDOFBUF To obtain the same behavior in NED, is enough to delete the message via 'no message' command.
The device will trim the trailing spaces, so if the user sets the follwing banner:
the device will delete the last spaces and the result will be:
In this case the compare-config diff is expected. Please be aware to trim the trailing spaces when the pre-login-banner is set via ncs_cli.
Special characters must be escaped. Example:
Please check below examples of how to configure a 'pre-login banner':
'\r\n' is missing from the command so no new line is provided
set a pre-login-banner on multiline
5. Built in live-status actions
Execution of the generic RPC
A. The generic RPC handling mechanism which is based on a pre-defined YANG model is designed to support any kind of device RPC. The user can send the commands to the device using a generic list of actions.Each action can be simple, interactive or internal. Please see below a general example for exporting the current configuration of the device:
To 'translate' the above CLI commands, the user can call the following list of actions:
Explanations:
list action: defines a list of RPC's to be sent to the device. Each action can be simple (like the first 2) or interactive (like the 'export-config' action)
action-payload: the actual CLI command sent to the device, e.g "show configuration" or "scope ssa" list interaction: used for interactive RPC's. Each entry defines a prompt-pattern and its corresponding value (e.g Password: -> password).
Notes:
prompt-pattern string is a regular expression, so special characters needs to be escaped;
the order of pattern definition is not important, so the list entries does not have to respect the device order of prompting.
The user will send a list of simple actions to go in the particular mode: 'scope ssa', 'scope system' and for the 'export-config' action, will need an interactive action to be able to send the password when the prompt 'Password:' will pop up. The regex 'Password.*' was chosen here to map the prompt Password:
Once the export-config was added, the user can call the 'show export-config' action to check the status as below:
To modify the field description, the user has to run the following commands:
The equivalent actions are:
Here the " are escaped using '\'.
In order to delete the 'export-config localhost' the user must issue the following command:
The related actions are:
B. Some of the generic RPC issued towards the device have a different prompt at the end of the execution. For example, suppose the prompt has the following format: 'prompt1# ' After a RPC is executed, the prompt will change. Example:
Also, multiple RPCs can be executed sequentially and the prompt may change one after another. For this situation, the user must configure a list of "expect-patterns" that comprises all the possible prompt patterns that can be met when a RPC is sent to the device. All the RPC will be treated as simple commands. To configure the list of "expect-patterns", the following setting should be done under ned-settings:
Please note that the value(".*> ?") for the list-entry above is a regex. If multiple regex are required to met all the possible prompt patterns, then configure multiple values for the list "expect-patterns".
Example of RPCs for a 93xx fxos device:
"connect module 1 telnet"
"connect ftd"
"system support diagnostic-cli"
"enable" (expects a "Password:" prompt" and a password with value: "test")
"show conn count"
The prompt is changing after every RPC but the common thing is that it ends with ">" or "> " or "text> ".
Hence, to match all the possibilities for these prompts, the expect-pattern will look like:
The interactive commands are the commands that expect a user interaction, like sending a username or password. RPC no 4, "enable" from above, is such an example.
The generic RPC command will be similar with:
In order to execute a new action in the same session, we have to leave the device in the init state, i.e. having the prompt from the beginning (when first connecting to it ). Hence, a few "exit" commands must be issued at the end of the generic RPCs.
Note: when a command is sent to the device, the echo of that command is expected. If for some reason, the echo is missing or is incomplete, then after X seconds (X is the number of seconds set for the connect-timeout), the following command from the RPCs set is sent to the device.
6. Built in live-status show
Examples of running live-status commands:
live-status command for monitoring Logical Devices: "show resource-profile user-defined"
live-status command for show interface under fabric a (if fabric b is supported, then interface-fabric-b can be used also).
live-status command for show slot and show monitor detail:
The time the values are stored in the memory is called ttl(time to live) and this in configurable under ned-settings:
50 represents the value of 50 seconds.
Example of running 'show' exec action:
7. Limitations
NED doesn't provide an exact copy of all details in the device CLI. Due to some YANG constraints, some of the CLI configuration commands may have a slightly different name.
Examples of CLI commands(differences between the device and NED ncs_cli)
8. How to report NED issues and feature requests
Issues like bugs and errors shall always be reported to the Cisco NSO NED team through the Cisco Support channel:
The following information is required for the Cisco NSO NED team to be able to investigate an issue:
Do as follows to gather the necessary information needed for your device, here named 'dev-1':
Enable full debug logging in the NED
Configure the NSO to generate a raw trace file from the NED
If the NED already had trace enabled, clear it in order to submit only relevant information
Do as follows for NSO 6.4 or newer:
Do as follows for older NSO versions:
Run a compare-config to populate the trace with initial device config
Reproduce the found issue using ncs_cli or your NSO service. Write down each necessary step in a reproduction report.
In addition to this, it helps if you can show how it should work by manually logging into the device using SSH/TELNET and type the relevant commands showing a successful operation.
Gather the reproduction report and a copy of the raw trace file containing data recorded when the issue happened.
Contact the Cisco support and request to open a case. Provide the gathered files together with access details for a device that can be used by the Cisco NSO NED when investigating the issue.
Requests for new features and extensions of the NED are handled by the Cisco NSO NED team when applicable. Such requests shall also go through the Cisco support channel.
The following information is required for feature requests and extensions:
Set the config on the real device including all existing dependent config and run sync-from to show it in the trace.
Run sync-from # devices device dev-1 sync-from
Attach the raw trace to the ticket
List the config you want implemented in the same syntax as shown on the device
SSH/TELNET access to a device that can be used by the Cisco NSO NED team for testing and verification of the new feature. This usually means that both read and write permissions are required. Pseudo access via tools like Webex, Zoom etc is not acceptable. However, it is ok with access through VPNs, jump servers etc as long as we can connect to the NED via SSH/TELNET.
9. How to rebuild a NED
To rebuild the NED do as follows:
When the NED has been successfully rebuilt, it is necessary to reload the package into NSO.
10. Configure the NED to use ssh multi factor authentication
This NED supports multi factor authentication (MFA) using the ssh authentication method 'keyboard-interactive'.
Some additional steps are required to enable the MFA support:
Verify that your NSO version supports MFA. This is configurable as additional settings in the authentication group used by the device instance.
Enter a NSO CLI and enter the following and do tab completion:
If 'mfa' is displayed in the output like above, NSO has MFA support enabled. In case MFA is not supported it is necessary to upgrade NSO before proceeding.
Implement the authenticator executable. The MFA feature relies on an external executable to take care of the client part of the multi factor authentication. The NED will automatically call this executable for each challenge presented by the ssh server and expects to get a proper response in return.
The executable can be a simple shell script or a program implemented in any programming language.
The required behaviour is like this:
read one line from stdin The line passed from the NED will be a semi colon separated string containing the following info:
The elements for device name, user, password and opaque corresponds to what has been configured in NSO. The ssh server name, instruction and prompt are given by the ssh server during the authentication step.
Each individual element in the semi colon separated list is Base64 encoded.
Extract the challenge based on the contents above.
Print a response matching the challenge to stdout and exit with code 0
In case a matching response can not be given do exit with code 2
Below is a simple example of an MFA authenticator implemented in Python3:
Configure the authentication group used by the device instance to enable MFA. There are two configurables available:
executable The path to the external multi factor authentication executable (mandatory).
opaque Opaque data that will passed as a cookie element to the executable (optional).
Try connecting to the device.
10.1 Trouble shooting
In case of connection problems the following steps can help for debugging:
Enable the NED trace in debug level:
Try connect again
Inspect the generated trace file.
Verify that the ssh client is using the external authenticator executable:
Verify that the executable is called with the challenges presented by the ssh server:
Check for any errors reported by the NED when calling the executable
11. Mandatory commands that have to be provided by the user
In order to have a perfect synchronization between the NED (and its CDB) and the target device, some commands that have default values on the device, must be provided by the user.
Please see the following examples:
enable/disable commands for the 'port-channel' or for the 'member-port' configured under the 'port-channel'. Creating port-channel 4 under eth-uplink/fabric a from ncs_cli:
If you would like to check how the commands that are going to be sent towards the device would look like, use the below dry-run command:
For the member-port:
When an interface becomes part of a member-port (interface 2 6 in the example below), the interface will be automatically removed by the device. Hence, to be sure the NED will be in sync with the device, the user must delete the related interface (interface 2 6) from fabric a. Example:
Note: this command is filtered and is not sent to the device but is used for keeping the internal CDB in sync with the dynamic configuration of the device.
When creating an external-port-link under the logical-device, a port-name command will be dynamically created on the device. The port-name command has to be provided in ncs_cli as follows:
On the device, the "interface Ethernet1/8" and "interface 1 8 have" the same meaning and the device saves the interfaces configuration using the last format. Hence, in the NED when you want to address the interface Ethernetx/y just use interface x y. (The transition to the interface Ethernetx/y will be handled internally, when necessary).
12. "security" section commands
local-user When creating a "local-user", the user must provide mandatory fields like: "password" and "phone" number. If the "phone" number is not provided, the fxos 21xx device will automatically set the phone number to
""value. Normally, the""is not allowed and the following operation is not permitted:set phone "". The fields with default values must be provided by the user, to maintain the synchronization between the NED and the device. The fxos 21xx device will automatically create "maxfailedlogins 0", so must be provided by the user as well.Example of creating a "local-user":
password-profile "change-interval" and "no-change-interval" cannot be both disabled at the same time.
fault policy/clear-interval never When "never" value is set for "clear-interval", the fxos device will automatically add one 0 for days, hours, minutes and seconds. Hence, "set clear-interval never" will be translated on the device "set clear-interval never 0 0 0 0". The user must provide all zero values and NED will filter them out.
13. How to configure additional config warning exceptions
In some situations, when configuring a particular command, the NED will treat the replies coming after, as an error if they are not part of known replies:
If one of the above reply is received by the NED, the command is considered correct, because it maches the known replies. If more similar replies are encountered, the NED list of known harmless replies must be updated. However, the user can get the same result by configuring "cisco-fxos write config-warning" ned-setting.
The 'config-warning' list key is a regular expression with a warning that should be ignored.
For example, if we receive a reply, which is splitted on more than 1 row and it doesn't match the known replies list from above, we can ignore it as below.
User can add a new warning that should be ignored, instead of throwing an error. Example:
when enabling "cc-mode" on the device, it returns a reply on 2 rows with the following content:
device-prompt # enable cc-mode
Reply from device: Warning: Connectivity to one or more services may be denied when committed. Please consult the product's CC Security Policy documentation. WARNING: A reboot of the system is required in order for the system to be operating in a CC approved mode.
This reply will lead to a NED error, since is not part of the known replies and it splitted on more than 1 row.
To avoid this, the user must add a new warning in the 'config-warning' list as below:
Notes:
in order for the warning exception to take effect, the ned-settings must be read again, using "disconnect" and "connect" commands.
the warning from the example above is already handled in NED and there is no need to be added under the "cisco-fxos write config-warning" section.
Last updated
Was this helpful?

