Looking for any experience from using Ansible to manage NSO runtime

[camoberg]

I believe I have heard mentions of using Ansible to manage the NSO development and runtime environments, and am interested to capture any learnings from that. This would include tasks like installing and upgrading of the NSO runtime environment and packages, and perhaps even using Ansible to manage the lifecycle of clusters or LSO deployments across servers.

[spudukko]

Part of branchInfra we have some basic flow , where we use ansible for nso and package installation. It is just a basic framework .

[camoberg]

Very cool. Do you have any playbooks that you would be willing to share?

[scott.barvick]

We have used Ansible for deployment of NSO for several projects.   We started to use it because a customer was using an NCS version that didn't support fully nct yet so we built up a set of playbooks that did everything from installing the packages to the target server to changing the permissions and groups on the NSO files in the system to reloading the packages on a running system.

Some customers, especially those on the latest NSO releases will upgrade NSO with each new release so Ansible is great for that.   Those who have a longer validation cycle for NSO do not want the playbook that upgrades NSO because it might be used

We use it to install NSO in customer environments and to local VMs for testing.   It is as simple as changing the inventory file that lists the hosts that that should get the deployment.

Because Ansible is just a wrapper around scripts with some special helpers, it can be adapted to do just about anything and frequently you will find that the Ansible folks have build in a function so that you don't have to do it with pure scripting e.g. here is a playbook for uploading packages to the NSO server.   The comments document the behavior and it is clear that Ansible has built in rsync, unzip, and a (somewhat odd syntax) file delete capability.  

---

  - name: Copy NSO and supporting packages and resources to the target

    hosts: all

    tasks:

          #

          # Synchronize the local distribution dir with the remote distribution dir

          # This will pick up everything in distribution (including packages) and

          # make sure it is on the target.  The synchronize command is a wrapper

          # around rsync so it is pretty efficient about only sending when things change

          #

          - name: Synchronize the target distribution directory with the local

            synchronize: src={{pkgsrc}} dest={{pkgdst}} recursive=yes delete=yes

          - name: Copy the resources to the target and expand

            unarchive: src={{pkgsrc}}/resources.tar.gz dest={{pkgdst}}/distribution

          - name: Delete the .tar.gz on the target

            file: path={{pkgdst}}/distribution/resources.tar.gz state=absent

[camoberg]

Scott,

Very cool, this is exactly what I was looking for. Would you be interested in contributing some of what you have to a devnet github repo and see if we can build a reusable, shareable set?

[frlundbe]

love it

[scott.barvick]

We can contribute what we have as a base.  I'll make it a bit more generic and submit it to github within a week or so.  Just like the other other examples we have seen so far, Ansible is very flexible and can do just about anything several different ways.

[camoberg]

Sorry for late reply, but: great!

[vleijon]

NGENA also uses ansible for deployments. I think it'd be great to standardize on some basic playbooks, especially if someone who is better than I am at ansible can make it fairly modular, it sounds like Scott's work is a good start.

[Robert Moss]

After deploying NSO in a fresh install via Ansible, I needed to run some NSO commands to set up Quagga and NSO HA components.

In order to do this, I wrote all the commands out to a file using Ansible blockinfile with variables, then executed ncs_cli as the nsoadmin (ncsadmin) user which ran the commands.

Example as follows:

  - name: create NSO command file

    blockinfile:

      path: "{{ nso_commands_file }}"

      create: yes

      mode: "0600"

      owner: "{{nso_user}}"    

      marker: "! {mark} Ansible section"

      block: |

        packages reload

        conf t

        devices template hcc-master config quagga-bgp:hostname MASTER

        devices template hcc-slave config quagga-bgp:hostname SLAVE

        devices template hcc-failover-master config quagga-bgp:hostname MASTER

        devices template hcc-none config quagga-bgp:hostname {{nso_ha_role}}

        commit

        !

        ha bgp anycast-path-min 2 anycast-prefix {{nso_ha_anycast_ip}}/{{nso_ha_anycast_ip_bits}}

        ha token {{rr_bgp_pass}} local-user {{nso_user}}

        commit

        !

        ha failure-limit 10

        ha interval 4

        ha member {{nso_ha_a_hostname}} address {{nso_ha_a_mgt_ip}} default-ha-role master quagga-device nso1-quagga cluster-manager true

        ha member {{nso_ha_b_hostname}} address {{nso_ha_b_mgt_ip}} default-ha-role slave quagga-device nso2-quagga failover-master true

        commit

        !

        exit

        exit

        ha commands activate

  # You can use shell to run other executables to perform actions inline

  - name: Run ncs_cli and load in devices

    command: /opt/nso/current/bin/ncs_cli -C {{ nso_commands_file }}

    become_user: "{{nso_user}}"

    become: true

  # clean up temp file

  - name: Remove NSO temp file

    file: path="{{ nso_commands_file }}" state=absent

    when: debug == 0