Setting Up Junos Node Slicing

Configuring an MX Series Router to Operate in BSYS Mode (External Server Model)

Junos node slicing requires the MX Series router to function as the base system (BSYS).

1. Install the Junos OS package for BSYS on both the Routing Engines of the MX Series router.

   To download the package:

   1. Go to the Juniper Support page.

   2. Click Base System > Junos OS version number > Junos version number (64-bit High-End).

   3. On the Software Download page, select the I Agree option under End User License Agreement and then click Proceed.

2. On the MX Series router, run the show chassis hardware command and verify that the transceivers on both the Control Boards (CBs) are detected. The following text represents a sample output:
   root@router> show chassis hardware
3. On the MX Series router, apply the following configuration statements:
   root@router# set chassis network-slices guest-network-functions

   root@router# set chassis redundancy graceful-switchover

   root@router# set chassis network-services enhanced-ip

   root@router# set routing-options nonstop-routing

   root@router# set system commit synchronize

   root@router# commit
   Note

   On MX960 routers, you must configure the network-services mode as enhanced-ip or enhanced-ethernet. On MX2020 routers, the enhanced-ip configuration statement is already enabled by default .

   The router now operates in BSYS mode.

Installing JDM RPM Package on x86 Servers Running RHEL (External Server Model)

Before installing the JDM RPM package for x86 servers, ensure that you have installed the additional packages, as described in Installing Additional Packages for JDM.

Download and install the JDM RPM package for x86 servers running RHEL as follows:

To download the package:

1. Go to the Juniper Support page.

2. Click JDM > Junos OS version number > Juniper Device Manager version number (for Redhat).

3. On the Software Download page, select the I Agree option under the End User License Agreement and then click Proceed.

1. Disable SELINUX and reboot the server. You can disable SELINUX by setting the value for SELINUX to disabled in the /etc/selinux/config file.
2. Install the JDM RPM package (indicated by the .rpm extension) by using the following command. An example of the JDM RPM package used is shown below:

   root@Linux Server0# rpm -ivh jns-jdm-1.0-0-17.4R1.13.x86_64.rpm

Repeat the steps for the second server.

Installing JDM Ubuntu Package on x86 Servers Running Ubuntu 16.04 (External Server Model)

Before installing the JDM Ubuntu package for x86 servers, ensure that you have installed the additional packages. For more details, see Installing Additional Packages for JDM.

Download and install the JDM Ubuntu package for x86 servers running Ubuntu 16.04 as follows:

To download the JDM Ubuntu package:

1. Go to the Juniper Support page.

2. Click JDM > Junos OS version number > Juniper Device Manager version number (for Debian).

3. On the Software Download page, select the I Agree option under the End User License Agreement and then click Proceed.

1. Disable apparmor and reboot the server.

   root@Linux Server0# systemctl stop apparmor

   root@Linux Server0# systemctl disable apparmor

   root@Linux Server0# reboot

2. Install the JDM Ubuntu package (indicated by the .deb extension) by using the following command. An example of the JDM Ubuntu package used is shown below:

Repeat the steps for the second server.

Configuring JDM on the x86 Servers (External Server Model)

1. At each server, start the JDM, and assign identities for the two servers as server0 and server1, respectively, as follows:

   On one server, run the following command:

   root@Linux server0# jdm start server=0

   On the other server, run the following command:

   root@Linux server1# jdm start server=1

   Note

   The identities, once assigned, cannot be modified without uninstalling the JDM and then reinstalling it:

2. Enter the JDM console on each server by running the following command:

   root@Linux Server0# jdm console
3. Log in as the root user.
4. Enter the JDM CLI by running the following command:

   root@jdm% cli

   Note

   The JDM CLI is similar to the Junos OS CLI.

5. Set the root password for the JDM.

   root@jdm# set system root-authentication plain-text-password

   Note

   • The JDM root password must be the same on both the servers.

   • Starting in Junos OS Release 18.3R1, you can create non-root users in JDM. For more information, see Configuring Non-Root Users in JDM (Junos Node Slicing).

   • JDM installation blocks libvirt port access from outside the host.

6. Commit the changes:

   root@jdm# commit

7. Enter Ctrl-] to exit from the JDM console.
8. From the Linux host, run the ssh jdm command to log in to the JDM shell.

Configuring Non-Root Users in JDM (Junos Node Slicing)

1. Log in to JDM as a root user.
2. Define a user name and assign the user with a predefined login class.

   root@jdm# set system login user username class predefined-login-class

3. Set the password for the user.

   root@jdm# set system login user username authentication plain-text-password
4. Commit the changes.

   root@jdm# commit

Table 1 contains the predefined login classes that JDM supports for non-root users:

Configuring JDM interfaces (External Server Model)

• The two 10-Gbps server ports that are connected to the MX Series router.

• The server port to be used as the JDM management port.

• The server port to be used as the GNF management port.

Therefore, you need to identify the following on each server before starting the configuration of the ports:

• The server interfaces (for example, p3p1 and p3p2) that are connected to CB0 and CB1 on the MX Series router.

• The server interfaces (for example, em2 and em3) to be used for JDM management and GNF management.

For more information, see the figure Connecting the Servers and the Router.

1. On server0, apply the following configuration statements:
   root@jdm# set groups server0 server interfaces cb0 p3p1

   root@jdm# set groups server0 server interfaces cb1 p3p2

   root@jdm# set groups server1 server interfaces cb0 p3p1

   root@jdm# set groups server1 server interfaces cb1 p3p2

   root@jdm# set apply-groups [ server0 server1 ]

   root@jdm# commit
   root@jdm# set groups server0 server interfaces jdm-management em2

   root@jdm# set groups server0 server interfaces vnf-management em3

   root@jdm# set groups server1 server interfaces jdm-management em2

   root@jdm# set groups server1 server interfaces vnf-management em3

   root@jdm# commit
2. Repeat the step 1 on server1.Note

   Ensure that you apply the same configuration on both server0 and server1.

3. Share the ssh identities between the two x86 servers.

   At both server0 and server1, run the following JDM CLI command:

   root@jdm> request server authenticate-peer-server

   Note

   The request server authenticate-peer-server command displays a CLI message requesting you to log in to the peer server using ssh to verify the operation. To log in to the peer server, you need to prefix ip netns exec jdm_nv_ns to ssh root@jdm-server1.

   For example, to log in to the peer server from server0, exit the JDM CLI, and use the following command from JDM shell:

   root@jdm:~# ip netns exec jdm_nv_ns ssh root@jdm-server1

   Similarly, to log in to the peer server from server1, use the following command:

   root@jdm:~# ip netns exec jdm_nv_ns ssh root@jdm-server0
4. Apply the configuration statements in the JDM CLI configuration mode to set the JDM management IP address, default route, and the JDM hostname for each JDM instance as shown in the following example.Note

   • The management IP address and default route must be specific to your network.

   • JDM does not support IPv6, even though IPv6 addresses are themselves configurable.

   root@jdm# set groups server0 interfaces jmgmt0 unit 0 family inet address 10.216.105.112/21

   root@jdm# set groups server1 interfaces jmgmt0 unit 0 family inet address 10.216.105.113/21

   root@jdm# set groups server0 routing-options static route 0.0.0.0/0 next-hop 10.216.111.254

   root@jdm# set groups server1 routing-options static route 0.0.0.0/0 next-hop 10.216.111.254

   root@jdm# set groups server0 system host-name test-jdm-server0

   root@jdm# set groups server1 system host-name test-jdm-server1

   root@jdm# commit
   Note

   • jmgmt0 stands for the JDM management port. This is different from the Linux host management port. Both JDM and the Linux host management ports are independently accessible from the management network.

   • You must have done the ssh key exchange as described in the Step 3 before attempting the Step 4. If you attempt the Step 4 without completing the Step 3, the system displays an error message as shown in the following example:

     Failed to fetch JDM software version from server1. If authentication of peer server is not done yet, try running request server authenticate-peer-server.

5. Run the following JDM CLI command on each server and ensure that all the interfaces are up.
   root@jdm> show server connections

If you want to modify the server interfaces configured in the JDM, perform the following steps:

1. Stop all running GNFs.
   root@jdm> request virtual-network-functions gnf-name stop
2. From the configuration mode, deactivate the virtual network functions configuration, and then commit the change.
   root@jdm# deactivate virtual-network-functions

   root@jdm# commit
3. Configure and commit the new interfaces as described in the step 1 of the main procedure.
4. Reboot the JDM from the shell.
   root@jdm:~# reboot
5. From the configuration mode, activate the virtual network functions configuration, and then commit the change.
   root@jdm# activate virtual-network-functions

   root@jdm# commit

Starting in Junos OS Release 19.2R1, Junos node slicing supports the assignment of a globally unique MAC address range (supplied by Juniper Networks) for GNFs. To know more, see Assigning MAC Addresses to GNF.

Configuring MX Series Router to Operate in In-Chassis Mode

In in-chassis model, the base system (BSYS), Juniper Device Manager (JDM), and all guest network functions (GNFs) run within the Routing Engine of the MX Series router. BSYS and GNFs run on the host as virtual machines (VMs). You need to first reduce the resource footprint of the standalone MX Series router as follows:

1. Ensure that both the Routing Engines (re0 and re1) in the MX Series router have the required VM host package (example: junos-vmhost-install-mx-x86-64-19.2R1.tgz) installed. The VM host package should be of 19.1R1 or a later version.
2. Applying the following configuration and then reboot VM host on both the Routing Engines (re0 and re1).
   user@router# set vmhost resize vjunos compact

   user@router# set system commit synchronize

   user@router> request vmhost reboot (re0|re1)

   When this configuration is applied, and following the reboot, the Routing Engine resource footprint of the Junos VM on MX Series router shrinks in order to accommodate GNF VMs. A resized Junos VM, now operating as the base system (BSYS) on the MX Series Routing Engine has the following resources:

   • CPU Cores—1 (Physical)

   • DRAM—16GB

   • Storage—14GB (/var)

Assigning MAC Addresses to GNF

1. root@jdm# set system vnf-license-supplement vnf-id gnf-id license-supplement-string augmented-ssrn-string

   root@jdm# commit
   Note

   • An augmented SSRN must be used for only one GNF ID. In the JDM, the GNF VMs are referred to as virtual network functions (VNFs). GNF ID is one of its attributes. Attributes of a VNF are fully described in the follow-on section Configuring Guest Network Functions.

   • By default, the augmented SSRN will be validated. Should you ever need to skip this validation, you can use the no-validate attribute in the CLI as follows: Example: set system vnf-license-supplement vnf-id gnf-id license-supplement-string augmented-ssrn-string [no-validate].

Configuring Guest Network Functions

Configuring a guest network function (GNF) comprises two tasks, one to be performed at the BSYS and the other at the JDM.

In the JDM, the GNF VMs are referred to as virtual network functions (VNFs). A VNF has the following attributes:

• A VNF name.

• A GNF ID. This ID must be the same as the GNF ID used at the BSYS.

• The MX Series platform type.

• A Junos OS image to be used for the GNF.

• The VNF server resource template.

At the JDM, to configure a VNF, perform the following steps:

1. Use the JDM shell command scp to retrieve the Junos OS Node Slicing image for GNF and place it in the JDM local directory /var/jdm-usr/gnf-images (repeat this step to retrieve the GNF configuration file).
   root@jdm:~# scp source-location-of-the-gnf-image /var/jdm-usr/gnf-images

   root@jdm:~# scp source-location-of-the-gnf-configuration-file /var/jdm-usr/gnf-config
2. Assign this image to a GNF by using the JDM CLI command as shown in the following example:
   root@test-jdm-server0> request virtual-network-functions test-gnf add-image /var/jdm-usr/gnf-images/junos-install-ns-mx-x86-64-17.4R1.10.tgz all-servers
3. Configure the VNF by applying the configuration statements as shown in the following example:
   root@test-jdm-server0# set virtual-network-functions test-gnf id 1

   root@test-jdm-server0# set virtual-network-functions test-gnf chassis-type mx2020

   root@test-jdm-server0# set virtual-network-functions test-gnf resource-template 2core-16g

   root@test-jdm-server0# set system vnf-license-supplement vnf-id test-gnf license-supplement-string RTU00023003204-01-AABBCCDDEE00-1100-01-411C

   For in-chassis model, do not configure the platform type (set virtual-network-functions test-gnf chassis-type mx2020). It will be detected automatically.

   Starting in Junos OS Release 19.2R1, Junos node slicing supports the assignment of a globally unique MAC address range (supplied by Juniper Networks) for GNFs. To know more, see Assigning MAC Addresses to GNF.

   To also specify a baseline or initial Junos OS configuration for a GNF, prepare the GNF configuration file (example: /var/jdm-usr/gnf-config/test-gnf.conf) on both the servers (server0 and server1) for external server model, and on both the Routing Engines (re0 and re1) for the in-chassis model, and specify the filename as the parameter in the base-config statement as shown below:

   root@test-jdm-server0# set virtual-network-functions test-gnf base-config /var/jdm-usr/gnf-config/test-gnf.conf

   root@test-jdm-server0# commit synchronize

   Note

   Ensure that:

   • You use the same GNF ID as the one specified earlier in BSYS.

   • The baseline configuration filename (with the path) is the same on both the servers / Routing Engines.

   • The syntax of the baseline file contents is in the Junos OS configuration format.

   • The GNF name used here is the same as the one assigned to the Junos OS image for GNF in the step 2.

4. To verify that the VNF is created, run the following JDM CLI command:

   root@test-jdm-server0> show virtual-network-functions test-gnf

5. Log in to the console of the VNF by issuing the following JDM CLI command:

   root@test-jdm-server0> request virtual-network-functions test-gnf console

6. Configure the VNF the same way as you configure an MX Series Routing Engine.

Chassis Configuration Hierarchy at BSYS and GNF

In Junos node slicing, the BSYS owns all the physical components of the router, including the line cards and fabric, while the GNFs maintain forwarding state on their respective line cards. In keeping with this split responsibility, Junos CLI configuration under the chassis hierarchy (if any), should be applied at the BSYS or at the GNF as follows:

• Physical-level parameters under the chassis configuration hierarchy should be applied at the BSYS. For example, the configuration for handling physical errors at an FPC is a physical-level parameter, and should therefore be applied at the BSYS.

  At BSYS Junos CLI:

  [edit]

  user@router# set chassis fpc fpc slot error major threshold threshold value action alarm

• Logical or feature-level parameters under the chassis configuration hierarchy should be applied at the GNF associated with the FPC. For example, the configuration for max-queues per line card is a logical-level parameter, and should therefore be applied at the GNF.

  At GNF Junos CLI:

  [edit]

  user@router# set chassis fpc fpc slot max-queues value

• As exceptions, the following two parameters under the chassis configuration hierarchy should be applied at both BSYS and GNF:

  At both BSYS and GNF CLI:

  [edit]

  user@router# set chassis network-services network services mode

  user@router# set chassis fpc fpc slot flexible-queueing-mode

Configuring Abstracted Fabric Interfaces Between a Pair of GNFs

Creating an Abstracted Fabric (af) interface between two guest network functions (GNFs) involves configurations both at the base system (BSYS) and at the GNF. Abstracted Fabric interfaces are created on GNFs based on the BSYS configuration, which is then sent to those GNFs.

1. At the BSYS, apply the configuration as shown in the following example:
   user@router# set chassis network-slices guest-network-functions gnf 2 af4 peer-gnf id 4

   user@router# set chassis network-slices guest-network-functions gnf 2 af4 peer-gnf af2

   user@router# set chassis network-slices guest-network-functions gnf 4 af2 peer-gnf id 2

   user@router# set chassis network-slices guest-network-functions gnf 4 af2 peer-gnf af4

   In this example, af2 is the Abstracted Fabric interface instance 2 and af4 is the Abstracted Fabric interface instance 4.

   Note

   The allowed af interface values range from af0 through af9.

   The GNF af interface will be visible and up. You can configure an af interface the way you configure any other interface.

2. At the GNF, apply the configuration as shown in the following example:
   user@router-gnf-b# set interfaces af4 unit 0 family inet address 10.10.10.1/24

   user@router-gnf-d# set interfaces af2 unit 0 family inet address 10.10.10.2/24

Class of Service on Abstracted Fabric Interfaces

Class of service (CoS) packet classification assigns an incoming packet to an output queue based on the packet’s forwarding class. See CoS Configuration Guide   for more details.

The following sections explain the forwarding class- to-queue mapping, and the behavior aggregate (BA) classifiers and rewrites supported on the Abstracted Fabric (af) interfaces.

• Forwarding Class-to-Queue Mapping

• BA Classifiers and Rewrites

Forwarding Class-to-Queue Mapping

An af interface is a simulated WAN interface with most capabilities of any other interface except that the traffic designated to a remote Packet Forwarding Engine will still have to go over the two fabric queues (Low/High priority ones).

Packets on the af interface inherit the fabric queue that is determined by the fabric priority configured for the forwarding class to which that packet belongs. For example, see the following forwarding class to queue map configuration:

[edit]

As shown in the preceding example, when a packet gets classified to the forwarding class VoiceSig, the code in the forwarding path examines the fabric priority of that forwarding class and decides which fabric queue to choose for this packet. In this case, high-priority fabric queue is chosen.

BA Classifiers and Rewrites

The behavior aggregate (BA) classifier maps a class-of-service (CoS) value to a forwarding class and loss priority. The forwarding class and loss-priority combination determines the CoS treatment given to the packet in the router. The following BA classifiers and rewrites are supported:

• Inet-Precedence classifier and rewrite

• DSCP classifier and rewrite

• MPLS EXP classifier and rewrite

  You can also apply rewrites for IP packets entering the MPLS tunnel and do a rewrite of both EXP and IPv4 type of service (ToS) bits. This approach will work as it does on other normal interfaces.

• DSCP v6 classifier and rewrite for IP v6 traffic

See CoS Configuration Guide   for details on CoS BA classifiers.

Optimizing Fabric Path for Abstracted Fabric Interface

To configure a fabric path optimization mode, use the following CLI commands at BSYS.

After configuring fabric path optimization, you can use the command show interfaces af-interface-name in GNF to view the number of packets that are currently flowing on the optimal / non-optimal path.

SNMP Trap Support: Configuring NMS Server (External Server Model)

The Juniper Device Manager (JDM) supports the following SNMP traps:

• LinkUp and linkDown traps for JDM interfaces.

  Standard linkUp/linkDown SNMP traps are generated. A default community string jdm is used.

• LinkUp/linkDown traps for host interfaces.

  Standard linkUp/linkDown SNMP traps are generated. A default community string host is used.

• JDM to JDM connectivity loss/regain traps.

  JDM to JDM connectivity loss/regain traps are sent using generic syslog traps (jnxSyslogTrap) through the host management interface.

  The JDM connectivity down trap JDM_JDM_LINK_DOWN is sent when the JDM is not able to communicate with the peer JDM on another server over cb0 or cb1 links. See the following example:

  The JDM to JDM Connectivity up trap JDM_JDM_LINK_UP is sent when either the cb0 or cb1 link comes up, and JDMs on both the servers are able to communicate again. See the following example:

• VM(GNF) up/down—libvirtGuestNotif notifications.

  For GNF start/shutdown events, the standard libvirtGuestNotif notifications are generated. For libvirtMIB notification details, see this web page. Also, see the following example:

SNMP traps are sent to the target NMS server. To configure the target NMS server details in the JDM, see the following example:

Sample Configuration for Junos Node Slicing

This section provides sample configurations for Junos node slicing.

Sample JDM Configuration (External Server Model)

Sample JDM Configuration (In-Chassis Model)

Sample BSYS Configuration with Abstracted Fabric Interface

Sample Abstracted Fabric Configuration at GNF with Class of Service

Assume that there is an Abstracted Fabric (af) interface between GNF1 and GNF2. The following sample configuration illustrates how to apply rewrites on the af interface at GNF1 and apply classifiers on the af interface on GNF2, in a scenario where traffic comes from GNF1 to GNF2:

GNF1 Configuration

GNF2 Configuration

Sample Output for Abstracted Fabric Interface State at a GNF