What triggers the prompt to ask server?

[spikefishjohn]

I have 3 gns3 boxes setup. I connect to the main and the other 2 are configured as remote servers.

If I drop a builtin deivces (switch or cloud) i'm asked which server do I want this to go on?

For any custom template it simply gets dropped on the main server. What do I need to do to get GNS3 to prompt for the custom devices (mostly qemu if that matters)?

[eantowne]

@spikefishjohn,

Assuming that each compute had the same custom template imported, there is one major problem. They cannot have the same name across multiple computes. In the current build, I do not believe there is a way to handle what you are looking for.

This would require some abstraction and a completely different concept for how templates and their respective images are handled.

[spikefishjohn]

Makes sad ansible templating syncing code sounds

@grossmj Any input on this? I was in the middle of writing a ansible script to duplicate templates from the main node to the remotes.

Having x*remotes templates for a single template doesn't seem optimal. I'd almost say they aren't worth using with remote servers if that's the only case.

Unless I'm thinking about this wrong.. can I use the template thats on the main server do deploy a VM on a remote by changing the compute_id field? I'm looking up the endpoint APIs so i might answer this myself shortly.

[grossmj]

Correct, the templates are on the controller and the compute_id is where a node based on the template will be created.

[spikefishjohn]

That didn't work, see below. I see this note on the template endpoint API

https://gns3-server.readthedocs.io/en/stable/api/v2/controller/template/projectsprojectidtemplatestemplateid.html

Name	Mandatory	Type	Description
compute_id		['null', 'string']	If the template don't have a default compute use this compute

I tried removing the compute_id from a template a few times and this made GNS3 very angry. I also tried settting compute_id to false. I'll poke around with that and see what happens.

This is what happens when I deploy a template located on compute01 with a target compute_id of "compute02".

gns3@compute01:~/GNS3/projects$ curl --no-progress-meter 'http://127.0.0.1:3080/v2/computes' | egrep name|compute_id
        compute_id: local,
        name: compute01,
        compute_id: 434e7036-e53e-459f-8349-bc160353b4b9,
        name: backup,
        compute_id: 7ef6b499-b732-40e4-b283-5b6d158e2ed0,
        name: compute02,
gns3@compute01:~/GNS3/projects$
gns3@compute01:~/GNS3/projects$ curl -X POST --no-progress-meter 'http://127.0.0.1:3080/v2/projects/a9e814b9-6572-4d3a-8e84-7fc63cd109d1/templates/c37e1c8d-d7dc-4c41-989e-67b36396f29f' -d '{x: 12, y: 42,compute_id: 7ef6b499-b732-40e4-b283-5b6d158e2ed0}' | egrep compute_id
    compute_id: local,
gns3@compute01:~/GNS3/projects$

[spikefishjohn]

Maybe this should be a bug report?

I created a template based on a switch and set compute_id to null (not "null") and then tried to send compute_id. After stop/restart of server and gui i'm now asked which server do I want the template on. After picking a server I get

Error while creating node from template: Compute ID None doesn't exist

If I try from curl I get the same error, so i think the but is with the template endpoint API.

gns3@compute01:~/.config/GNS3/2.2$ curl -X POST  'http://127.0.0.1:3080/v2/projects/cb65dac2-ad5b-47b0-bc3c-08db2991f1b3/templates/b43f7082-57e2-4363-ae41-6a4714b3509e' -d '{"x": 12, "y": 42, "compute_id": "7ef6b499-b732-40e4-b283-5b6d158e2ed0"}'
{
    "message": "Compute ID None doesn't exist",
    "status": 404
}
gns3@compute01:~/.config/GNS3/2.2$

Here is the template I created. I'm not showing all the ports, i just figured 7 or so wasn't useful.

    {
        "builtin": false,
        "category": "switch",
        "compute_id": null,
        "console_type": "none",
        "default_name_format": "Switch{0}",
        "name": "Ethernet switchSpec",
        "ports_mapping": [
            {
                "ethertype": "",
                "name": "Ethernet0",
                "port_number": 0,
                "type": "access",
                "vlan": 1
            },

        ],
        "symbol": ":/symbols/ethernet_switch.svg",
        "template_id": "b43f7082-57e2-4363-ae41-6a4714b3509e",
        "template_type": "ethernet_switch"
    },

[spikefishjohn]

Ok, in my continued quest for this I created a pull request. This makes the /v2/project/{project_id}/template/{template_id} api call accept compute_id when passed in a json request to drop a new node.

I also updated api template handler to not remove compute_id from an update request to /v2/template/{template_id}. This way I can convert all my custom template compute IDs to null. The GUI is now asking me which server I want to drop any template on.

NOTE: This also breaks the fat client's ability to see disk images when configuring a template since all my templates are null.

An interesting thing I just found out. If the main server detects the remote server is missing the hard drive image for a given template it will automagically upload it. Who knew!

Here is my pull request

GNS3/gns3-server#2077

Here is the ansible role code I used to update all my templates (roles/update-compute-ids-to-null/tasks/main.yml). I'm using auth, if you aren't just comment out the url_password and url_username.

- name: Get all templates in json format
  uri:
    url: 'http://127.0.0.1:3080/v2/templates'
    url_password: "{{ gns3_password }}"
    url_username: "{{ gns3_username }}"
    return_content: yes
    body_format: json
    force_basic_auth: yes
  register: templates

- name: Loop through all none builtin templates and set compute_id to null via merge.
  uri:
    url: "{{ 'http://127.0.0.1:3080/v2/templates/' + item.template_id }}"
    url_password: "{{ gns3_password }}"
    url_username: "{{ gns3_username }}"
    method: PUT
    body: "{{ item }}"
    force_basic_auth: yes
    body_format: json
  loop: "{{ templates.json | json_query('[*].merge(@, {compute_id: null}) | [?builtin == `false` ]') }}"