Revision as of 02:27, 22 October 2012

A Cisco Plugin Framework for Quantum Supporting L2 Networks Spannning Multiple Switches

README for Quantum v2.0: A Plugin Framework for Supporting Quantum Networks Spannning Multiple Switches

Introduction

This plugin implementation provides the following capabilities:

• A reference implementation for a Quantum Plugin Framework (For details see: http://wiki.openstack.org/quantum-multi-switch-plugin)
• Supports multiple switches in the network
• Supports multiple models of switches concurrently
• Supports use of multiple L2 technologies
• Supports the Cisco Nexus family of switches.
• Supports Cisco UCS blade servers with M81KR Virtual Interface Cards (aka "Palo adapters") via 802.1Qbh.

Pre-requisites

(The following are necessary only when using the UCS and/or Nexus devices in your system. If you plan to just leverage the plugin framework, you do not need these.)

If you are using a Nexus switch in your topology, you'll need the following NX-OS version and packages to enable Nexus support:

• NX-OS 5.2.1 (Delhi) Build 69 or above.
• paramiko library - SSHv2 protocol library for python
• ncclient v0.3.1 - Python library for NETCONF clients
  • You need a version of ncclient modified by Cisco Systems. To get it, from your shell prompt do:

git clone git@github.com:CiscoSystems/ncclient.git
sudo python ./setup.py install

• For more information of ncclient, see: http://schmizz.net/ncclient/
• One or more UCS B200 series blade servers with M81KR VIC (aka Palo adapters) installed.
• UCSM 2.0 (Capitola) Build 230 or above.
• OS supported:
  • RHEL 6.1 or above
  • Ubuntu 11.10 or above
  • Package: python-configobj-4.6.0-3.el6.noarch (or newer)
  • Package: python-routes-1.12.3-2.el6.noarch (or newer)
  • Package: pip install mysql-python

Module Structure

• quantum/plugins/cisco/ - Contains the Network Plugin Framework
  • /client - CLI module for core and extensions API
  • /common - Modules common to the entire plugin
  • /conf - All configuration files
  • /db - Persistence framework
  • /models - Class(es) which tie the logical abstractions to the physical topology
  • /nova - Scheduler and VIF-driver to be used by Nova
  • /nexus - Nexus-specific modules
  • /segmentation - Implementation of segmentation manager, e.g. VLAN Manager
  • /services - Set of orchestration libraries to insert In-path Networking Services
  • /tests - Tests specific to this plugin
  • /ucs - UCS-specific modules

Plugin Installation Instructions

1. Make a backup copy of quantum/etc/quantum.conf

2. Edit quantum/etc/quantum.conf and edit the "core_plugin" for v2 API

`core_plugin = quantum.plugins.cisco.network_plugin.PluginV2`

3. MySQL database setup:

   3a.  Create quantum_l2network database in mysql with the following command -

`mysql -u<mysqlusername> -p<mysqlpassword> -e "create database quantum_l2network"`

   3b.  Enter the quantum_l2network database configuration info in the
        quantum/plugins/cisco/conf/db_conn.ini file.

4. If you want to turn on support for Cisco Nexus switches:

   4a.  Uncomment the nexus_plugin property in
        etc/quantum/plugins/cisco/cisco_plugins.ini to read:

[PLUGINS]
nexus_plugin=quantum.plugins.cisco.nexus.cisco_nexus_plugin_v2.NexusPlugin

   4b.  Enter the relevant configuration in the
        etc/quantum/plugins/cisco/nexus.ini file.  Example:

[SWITCH]
# Change the following to reflect the IP address of the Nexus switch.
# This will be the address at which Quantum sends and receives configuration
# information via SSHv2.
nexus_ip_address=10.0.0.1
# Port numbers on the Nexus switch to each one of the compute nodes are connected
# Use shortened interface syntax, e.g. "1/10" not "Ethernet1/10" and "," between ports.
ports=1/10,1/11,1/12
#Port number where SSH will be running on the Nexus switch.  Typically this is 22
#unless you've configured your switch otherwise.
nexus_ssh_port=22

[DRIVER]
name=quantum.plugins.cisco.nexus.cisco_nexus_network_driver.CiscoNEXUSDriver

   4c.  Make sure that SSH host key of the Nexus switch is known to the
        host on which you are running the Quantum service.  You can do
        this simply by logging in to your Quantum host as the user that
        Quantum runs as and SSHing to the switch at least once.  If the
        host key changes (e.g. due to replacement of the supervisor or
        clearing of the SSH config on the switch), you may need to repeat
        this step and remove the old hostkey from ~/.ssh/known_hosts.

5. If your are using UCS blade servers with M81KR Virtual Interface Cards and

   want to leverage the VM-FEX features,

   5a.  Uncomment the ucs_plugin propertes in
        etc/quantum/plugins/cisco/cisco_plugins.ini to read:

[PLUGINS]
ucs_plugin=quantum.plugins.cisco.ucs.cisco_ucs_plugin_v2.UCSVICPlugin
[INVENTORY]
ucs_plugin=quantum.plugins.cisco.ucs.cisco_ucs_inventory_v2.UCSInventory

    5b.  Enter the relevant configuration in the
         etc/quantum/plugins/cisco/ucs.ini file.  Example:

[UCSM]
#change the following to the appropriate UCSM IP address
#if you have more than one UCSM, enter info from any one
ip_address=<put_ucsm_ip_address_here>
default_vlan_name=default
default_vlan_id=1
max_ucsm_port_profiles=1024
profile_name_prefix=q-

[DRIVER]
name=quantum.plugins.cisco.ucs.cisco_ucs_network_driver.CiscoUCSMDriver

   5c.  Configure the UCS systems' information in your deployment by editing the
        quantum/plugins/cisco/conf/ucs_inventory.ini file. You can configure multiple
        UCSMs per deployment, multiple chassis per UCSM, and multiple blades per
        chassis. Chassis ID and blade ID can be obtained from the UCSM (they will
        typically be numbers like 1, 2, 3, etc.). Also make sure that you put the exact
        hostname as nova sees it (the host column in the services table of the nova
        DB will give you that information).

[ucsm-1]
ip_address = <put_ucsm_ip_address_here>
[[chassis-1]]
chassis_id = <put_the_chassis_id_here>
[[[blade-1]]]
blade_id = <put_blade_id_here>
host_name = <put_hostname_here>
[[[blade-2]]]
blade_id = <put_blade_id_here>
host_name = <put_hostname_here>
[[[blade-3]]]
blade_id = <put_blade_id_here>
host_name = <put_hostname_here>

[ucsm-2]
ip_address = <put_ucsm_ip_address_here>
[[chassis-1]]
chassis_id = <put_the_chassis_id_here>
[[[blade-1]]]
blade_id = <put_blade_id_here>
host_name = <put_hostname_here>
[[[blade-2]]]
blade_id = <put_blade_id_here>
host_name = <put_hostname_here>

   5d. Configure your OpenStack installation to use the 802.1qbh VIF driver and
       Quantum-aware scheduler by editing the /etc/nova/nova.conf file with the
       following entries:

scheduler_driver=quantum.plugins.cisco.nova.quantum_port_aware_scheduler.QuantumPortAwareScheduler
quantum_host=127.0.0.1
quantum_port=9696
libvirt_vif_driver=quantum.plugins.cisco.nova.vifdirect.Libvirt802dot1QbhDriver
libvirt_vif_type=802.1Qbh

   Note: To be able to bring up a VM on a UCS blade, you should first create a
         port for that VM using the Quantum create port API. VM creation will
         fail if an unused port is not available. If you have configured your
         Nova project with more than one network, Nova will attempt to instantiate
         the VM with one network interface (VIF) per configured network. To provide
         plugin points for each of these VIFs, you will need to create multiple
         Quantum ports, one for each of the networks, prior to starting the VM.
         However, in this case you will need to use the Cisco multiport extension
         API instead of the Quantum create port API. More details on using the
         multiport extension follow in the section on multi NIC support.

   To support the above configuration, you will need some Quantum modules. It's easiest
   to copy the entire quantum directory from your quantum installation into:

   /usr/lib/python2.7/site-packages/

   This needs to be done on each nova compute node.

7. Verify that you have the correct credentials for each IP address listed

   in quantum/plugins/cisco/conf/credentials.ini.  Example:

# Provide the UCSM credentials, create a separte entry for each UCSM used in your system
# UCSM IP address, username and password.
[10.0.0.2]
username=admin
password=mySecretPasswordForUCSM

# Provide the Nexus credentials, if you are using Nexus switches.
# If not this will be ignored.
[10.0.0.1]
username=admin
password=mySecretPasswordForNexus

   In general, make sure that every UCSM and Nexus switch  used in your system,
   has a credential entry in the above file. This is required for the system to
   be able to communicate with those switches.

9. Start the Quantum service. If something doesn't work, verify the

   your configuration of each of the above files.

Multi NIC support for VMs

As indicated earlier, if your Nova setup has a project with more than one network, Nova will try to create a virtual network interface (VIF) on the VM for each of those

As indicated earlier, if your Nova setup has a project with more than one network, Nova will try to create a virtual network interface (VIF) on the VM for each of those networks. Before each VM is instantiated, you should create Quantum ports on each of those networks. These ports need to be created using the following rest call:

POST /1.0/extensions/csco/tenants/{tenant_id}/multiport/

with request body:

{'multiport':
 {'status': 'ACTIVE',
  'net_id_list': net_id_list,
  'ports_desc': {'key': 'value'

</nowiki></pre>

where,

net_id_list is a list of network IDs: [netid1, netid2, ...]. The "ports_desc" dictionary is reserved for later use. For now, the same structure in terms of the dictionary name, key and value should be used.

The corresponding CLI for this operation is as follows:

`PYTHONPATH=. python quantum/plugins/cisco/client/cli.py create_multiport <tenant_id> <net_id1,net_id2,...>`

   (Note that you should not be using the create port core API in the above case.)

Using an independent plugin as a device sub-plugin

If you would like to use an independent virtual switch plugin as one of the sub-plugins (for eg: the OpenVSwitch plugin) with the nexus device sub-plugin perform the following steps:

(The following instructions are with respect to the OpenVSwitch plugin) 1. Update etc/quantum/plugins/cisco/l2network_plugin.ini

  In the [MODEL] section of the configuration file put the following configuration
  (note that this should be the only configuration in this section, all other configuration
  should be either removed or commented)

   `model_class=quantum.plugins.cisco.models.virt_phy_sw_v2.VirtualPhysicalSwitchModelV2`

2. Update etc/quantum/plugins/cisco/cisco_plugins.ini

  In the [PLUGINS] section of the configuration file put the following configuration:

  `vswitch_plugin=quantum.plugins.openvswitch.ovs_quantum_plugin.OVSQuantumPluginV2`

3. Set the DB name, the same name has to be configured in three places:

  In etc/quantum/plugins/cisco/conf/db_conn.ini set the "name" value
  In /etc/quantum/plugins/openvswitch/ovs_quantum_plugin.ini set the "sql_connection"

  In etc/quantum/plugins/cisco/conf/db_conn.ini set the "name" value
  In /etc/quantum/plugins/openvswitch/ovs_quantum_plugin.ini set the "sql_connection"
  In /etc/quantum/dhcp_agent.ini set the "db_connection"

4. The range of VLAN IDs has to be set in the OpenVSwitch configuration file:

  In /etc/quantum/plugins/openvswitch/ovs_quantum_plugin.ini
  Set:

   vlan_min = <lower_id>
   vlan_max = <higher_id>
   enable_tunneling = False

5. For Nexus device sub-plugin configuration refer to the above sections

How to test the installation

The unit tests are located at quantum/plugins/cisco/tests/unit/v2. They can be executed from the top level Quantum directory using the run_tests.sh script.

1. Testing the core API (without UCS/Nexus/RHEL device sub-plugins configured):

  By default all the device sub-plugins are disabled (commented out) in
  etc/quantum/plugins/cisco/cisco_plugins.ini

   ./run_tests.sh quantum.plugins.cisco.tests.unit.v2.test_api_v2
   ./run_tests.sh quantum.plugins.cisco.tests.unit.v2.test_network_plugin

2. For testing the Nexus device sub-plugin perform the following configuration:

  Edit etc/quantum/plugins/cisco/cisco_plugins.ini to add:
  In the [PLUGINS] section add:

`nexus_plugin=quantum.plugins.cisco.nexus.cisco_nexus_plugin_v2.NexusPlugin`

  Edit the etc/quantum/plugins/cisco/nexus.ini file.
  When not using Nexus hardware use the following dummy configuration verbatim:

[SWITCH]
nexus_ip_address=1.1.1.1
ports=1/10,1/11,1/12
nexus_ssh_port=22
[DRIVER]
name=quantum.plugins.cisco.tests.unit.v2.nexus.fake_nexus_driver.CiscoNEXUSFakeDriver
   Or when using Nexus hardware (put the values relevant to your setup):
[SWITCH]
nexus_ip_address=1.1.1.1
ports=1/10,1/11,1/12
nexus_ssh_port=22
[DRIVER]
name=quantum.plugins.cisco.nexus.cisco_nexus_network_driver.CiscoNEXUSDriver

  (Note: Make sure that quantum/plugins/cisco/conf/credentials.ini has an entry for

3. For testing the UCS device sub-plugin perform the following configuration:

  Edit etc/quantum/plugins/cisco/cisco_plugins.ini to add:
  In the [PLUGINS] section add:

`ucs_plugin=quantum.plugins.cisco.ucs.cisco_ucs_plugin_v2.UCSVICPlugin`

  In the [INVENTORY] section add:
  When not using UCS hardware:

`ucs_plugin=quantum.plugins.cisco.tests.unit.v2.ucs.cisco_ucs_inventory_fake.UCSInventory`

  Or when using UCS hardware:

`ucs_plugin=quantum.plugins.cisco.ucs.cisco_ucs_inventory_v2.UCSInventory`

  Edit the etc/quantum/plugins/cisco/ucs.ini file.
  When not using UCS hardware:

[DRIVER]
name=quantum.plugins.cisco.tests.unit.v2.ucs.fake_ucs_driver.CiscoUCSMFakeDriver
   Or when using UCS hardware:
[DRIVER]
name=quantum.plugins.cisco.ucs.cisco_ucs_network_driver.CiscoUCSMDriver

Copyright: 2012 Cisco Systems, Inc.