The-Nazara-Project / Nazara

A CLI application to create and update machines and VMs in Netbox.
GNU General Public License v3.0
11 stars 2 forks source link

Implement linking of IP Adresses to newly created device. #73

Closed ByteOtter closed 2 months ago

ByteOtter commented 7 months ago

What does this PR change?

Currently we can only create devices without the IP address fields being set.

This PR aims to fix that. For this, after device creation, we need to create an Interface for the device, create an IP address object and then PATCH the device with the ID of its primary IP address linked to one of its devices.

As references serve these threads:

https://github.com/netbox-community/netbox/discussions/8746

https://github.com/netbox-community/netbox/discussions/8837

Tick the applicable box:

Links

Fixes: #69

Documentation

ByteOtter commented 7 months ago

Currently we do not get a valid JSON response body when the creation is successful. We need to investigate if this issue is on our end or on theirs. A possible workaround could look like this:

  1. Create interface only checking for valid errors and ignoring JSON parsing errors.
  2. Get list of all Interfaces from NetBox
  3. Search list for the interface we just created. Best guess would be the MAC-Address?

Problems with this workaround:

Edit: The problem seems to be with us as running it manually:

curl -X POST https://demo.netbox.dev/api/dcim/interfaces/ \
-H "Authorization: Token $TOKEN" \
-H "Content-Type: application/json" \
-d '{"name": "Nazara-Test", "device": 112, "type": "1000base-t"}' \
-o ~/Documents/test.txt

Returns what appears to be a valid JSON:

{
    "id":2494,
    "url":"https://demo.netbox.dev/api/dcim/interfaces/2494/",
    "display":"Nazara0",
    "device":{
        "id":112,"url":"https://demo.netbox.dev/api/dcim/devices/112/",
        "display":"CISCO",
        "name":"CISCO"
        },
    "vdcs":[],
    "module":null,
    "name":"Nazara0",
    "label":"",
    "type":
        {
        "value":"1000base-t",
        "label":"1000BASE-T (1GE)"
        },
    "enabled":true,
    "parent":null,
    "bridge":null,
    "lag":null,
    "mtu":null,
    "mac_address":null,
    "speed":null,
    "duplex":null,
    "wwn":null,
    "mgmt_only":false,
    "description":"",
    "mode":null,
    "rf_role":null,
    "rf_channel":null,
    "poe_mode":null,
    "poe_type":null,
    "rf_channel_frequency":null,
    "rf_channel_width":null,
    "tx_power":null,
    "untagged_vlan":null,
    "tagged_vlans":[],
    "mark_connected":false,
    "cable":null,
    "cable_end":"",
    "wireless_link":null,
    "link_peers":[],
    "link_peers_type":null,
    "wireless_lans":[],
    "vrf":null,
    "l2vpn_termination":null,
    "connected_endpoints":null,
    "connected_endpoints_type":null,
    "connected_endpoints_reachable":null,
    "tags":[],
    "custom_fields":{},
    "created":"2024-04-25T11:55:29.998975Z",
    "last_updated":"2024-04-25T11:55:29.999015Z",
    "count_ipaddresses":0,
    "count_fhrp_groups":0,
    "_occupied":false
}

This does not make sense though as we use serde to handle responses and other endpoints work flawlessly.

Edit Edit:

It looks like l2vpn_termination is null. This is fine for creation but cannot be parsed as the struct serde tries to make out of the response (DcimInterfacesCreateResponse) said that it is not an Option.

according to netbox's code this should be nullable. Meaning their spec is wrong...

l2vpn_termination = NestedL2VPNTerminationSerializer(read_only=True, allow_null=True)
ByteOtter commented 2 months ago

After further investigation, it definitely looks like this is a problem on NetBox's end. Looking at the fields link_peers_type and connected_endpoints in the Interface struct, we can see that the schema requires them to be a String and a Vec. Which gets translated correctly in thanix_client. However, the application returns both of these values as null, which does not correspond with the schema.

This leads to several conclusions:

  1. This could a problem that can re-occur in more fields than the ones we encountered here.
  2. A temporary workaround is required to allow v0.1.0 to be working with NetBox v3.6.x as per the decision made in #74.
  3. An automated integration test suite is required to catch these errors correctly.

Backend bug hunting must also be conducted as we cannot be sure this problem is fixed with NetBox v4.x. I had opened an issue with NetBox about a possible bug on this end (https://github.com/netbox-community/netbox/issues/16301) they asserted that this is not an issue with them. We need to collect further information before we can proceed to report a bug again.