Crash: AttributeError: 'dict' object has no attribute 'matches_site_cluster' when a VLAN's group is an unresolved reference

[tempalov]

Environment

• netbox-sync 1.8.0 (official bbricardo/netbox-sync:1.8.0 image) — the same code is still present on main
• NetBox 4.6.4
• Source type: vmware (8 vCenters)

Problem

Every run aborts during source.apply():

File "/app/module/sources/vmware/connection.py", line 2055, in add_host
    self.add_device_vm_to_inventory(NBDevice, object_data=host_data, pnic_data=pnic_data_dict,
File "/app/module/sources/vmware/connection.py", line 1186, in add_device_vm_to_inventory
    nic_object, ip_address_objects = self.add_update_interface(nic_object_dict.get(int_name), device_vm_object,
File "/app/module/sources/common/source_base.py", line 668, in add_update_interface
    matching_tagged_vlan = self.get_vlan_object_if_exists(tagged_vlan, device_object_site,
File "/app/module/sources/common/source_base.py", line 886, in get_vlan_object_if_exists
    if vlan_group.matches_site_cluster(vlan_site, vlan_cluster):
AttributeError: 'dict' object has no attribute 'matches_site_cluster'

Because the exception is raised inside source.apply(), it happens before update_instance(), so a failing run writes nothing to NetBox. In our case this silently froze all virtualization.* data for several weeks.

Cause

In get_vlan_object_if_exists():

# try find matching VLAN by group
if grab(vlan, "data.group") is not None:
    vlan_group = grab(vlan, "data.group")
    if vlan_group.matches_site_cluster(vlan_site, vlan_cluster):

the code assumes vlan.data["group"] is always a resolved NBVLANGroup instance. That holds only for VLAN groups that are part of the inventory netbox-sync fetched itself. For VLANs that already exist in NetBox and whose VLAN group was created by another tool, the relation is never resolved and stays a raw API dict:

{'id': 53392, 'url': '.../api/ipam/vlan-groups/53392/', 'display': '...', 'name': '...'}

The run then dies at the first ESXi host that has a tagged VLAN whose VID matches such a VLAN. The log shows a large number of Problems resolving relation ... messages beforehand, which is the same underlying condition.

How to reproduce

1. Have NetBox contain VLANs assigned to VLAN groups that netbox-sync does not manage. In our case ~350 groups were created by NetBox Labs orb-agent / Diode ingestion.
2. Run netbox-sync against a vmware source whose hosts carry tagged VLANs with VIDs matching those VLANs.

The trigger is purely data-side: the exact same netbox-sync version ran fine for months and started crashing only after those VLAN groups appeared.

Existing options do not help

disable_vlan_sync, vlan_sync_exclude_by_name and vlan_sync_exclude_by_id cannot work around this — all three are evaluated inside add_vlan_object_to_netbox(), which is called after get_vlan_object_if_exists() has already crashed.

Suggested fix

if isinstance(vlan_group, NetBoxObject) and vlan_group.matches_site_cluster(vlan_site, vlan_cluster):

If the group cannot be resolved to an object, no group match is attempted and the existing site / global VLAN matching applies — the correct degraded behaviour. NetBoxObject is already in scope via from module.netbox import *.

Verification

Tested against our production data set (8 vCenters, ~1300 VMs, NetBox with ~350 foreign VLAN groups):

	unpatched	patched
sources processed	0 of 8 — crashed on the first	8 of 8
time to crash	~8 s after the first source started	no crash, full run 1 h 16 min
AttributeError	reproduced on every run	0

Both runs crashed/passed on the identical trigger host, so the change is what makes the difference.

I will open a PR with this one-line change.

[semx]

Fixed in development by #530 (merged), with a regression test in #548. Thanks @tempalov for the precise diagnosis and the fix. Ships with the next release.