Gitiles
Code Review Sign In
gerrit.opencord.org / maas / e72d83d53a5adf9d1900c788b2c00595a49e0402
commite72d83d53a5adf9d1900c788b2c00595a49e0402[log] [tgz]
authorAndy Bavier <andy@onlab.us>Wed Jun 14 17:28:30 2017 -0700
committerAndy Bavier <andy@opennetworking.org>Fri Sep 08 10:57:54 2017 -0700
tree8eb2076f34e2aa202a56b15b0d913951d53e965a
parente2d8db44abcf5c0f79c1ab6f24288868a8890f93 [diff]
CORD-1283 Workaround for Calix pod install failure

Change-Id: I74950cd0e16e3b69e1033f394b71b9c7b078dba9
(cherry picked from commit ad44c16d3322b4e89fb7505f7be499a86e36b928)
2 files changed
tree: 8eb2076f34e2aa202a56b15b0d913951d53e965a
  1. .gitignore
  2. API.md
  3. HINTS.md
  4. LICENSE.txt
  5. Makefile
  6. README.md
  7. Vagrantfile
  8. ansible.cfg
  9. ansible/
  10. automation/
  11. build.gradle
  12. buildSrc/
  13. compute-node.yml
  14. compute.yml
  15. config-generator/
  16. deploy-vars
  17. dev-head-node.yml
  18. doc/
  19. fabric.yml
  20. gradle.properties
  21. gradle/
  22. gradlew
  23. gradlew.bat
  24. group_vars/
  25. harvester/
  26. head-node.yml
  27. head.yml
  28. help.mk
  29. host_vars/
  30. hosts
  31. ip-allocator/
  32. library/
  33. prime-node.yml
  34. provisioner/
  35. roles/
  36. rules.mk
  37. scripts/
  38. setup-ciab-pcu.yml
  39. switchq/
  40. utils/
README.md

OpenCORD Bare Metal Provisioning

OpenCORD leverages Canonical's Metal as a Service (MAAS) solution. The MAAS solution provides a PXE boot environment. The basic bare metal provisioning flow is:

  1. Install and provisioning MAAS and other utilities on one compute node that will have the head node role
  2. Boot the other components in a CORD POD (switches and other compute nodes)
  3. Once other components are operational perform some additional provisioning to prepare them to be part of a CORD POD and to be compliant with best practices of a CORD POD

After the base bare metal provisioning is complete further provisioning, such as XOS or leaf-spine fabric can be deployed.

Ansible Roles Provided

docker

Ensures Docker tools are available on the target system. Specifically docker-engine and docker-compose.

java8-oracle

Ensures that the Oracle version of Java8 is available on the target system.

fabric-switch

Ensures the OpenFlow agent (ofdpa) is available on the target switch as well as utility scripts that have been helpful in the lab.

Configuration

One of the scripts made available via this role is called connect. This script initiates the connection from the switch to an OpenFlow controller. To connect to the controller the DPID of the switch is required as is the IP Address of the SDN controller.

The IP Address of the controller is likely universal for all switches and can be either set via the command line, using the --extra-vars command line option or via a global vars file.

The DPID is a per switch setting and represents the OpenFlow ID for the switch and will be in the for or 0x0000000000000012 and will be unique for each switch. This value can be set either in a host specific variable file or if the playbook is being run against a single switch using the --extra-vars command line option.

compute-node

Ensures the 40G network interface card (NIC) drivers are installed and that the interfaces on the compute node are named according to best practices. Specificall the 2 40G ports are eth0 and eth1; the 2 10G ports are eth2 and eth3. Additionally, this roles sets a default password for the ubuntu user so that console logins are possible for debug purposes. This last change, default password, should be eliminated for proxuction use.

Configuration

Each compute node is statically assigned an IP address for the leaf-spine fabric. This can be configured in a host specific variable file found in the host_vars directory or could be specified on the command line using the --extra-vars option if the play book is be run against a single target system.

Dependencies

This role depends on the docker role.

onos-fabric

Ensures

mmas

Ensures that Canonical's Metal as a Service (MAAS) is available on the target system and configured according to best practices for a CORD POD. This role is meant to be applied to a head node in the CORD POD.

Assumptions

  • A 2 port Intel 40G card is installed on the head node
  • A 2 port 10G card is installed on the head node
  • Head Node has Internet connectivity via the 10G interface named eth3

Additionally configuration variables, including network IP addressing information can be found in the file vars/main.yml.

Configuration

- When working with virtual machines, or really any machine that does not have a real 40G interface it is
  sometimes useful to not attempt to rename / reconfigure the machines interfaces. This can be accomplished
  by adding the `--skip-tags=interface_config` command line option to your ansible-playbook command line.

Comments

  • iptables rules will be established to NAT traffic out interface eth3
  • This role installs and starts two docker images to help manage the MAAs install including:
    • cord/maas-dhcp-harvester:0.1-prerelease - adds DHCP addresses to the DNS server that MAAS misses
    • cord/maas-automation:0.1-prerelease - automates compute nodes through the states of MAAS to the deployed state so that they can use used as part of the CORD POD.

Dependencies

This role depends on the compute-node role.