fix(sync): handle 'already exists' race condition in try_create

[berlikm]

Summary

When multiple RQ workers sync devices concurrently, two jobs can both pass (returns empty), then both call for the same hostgroup/template/tag. Zabbix rejects the second with "already exists". This propagates as a RuntimeError and aborts the entire host sync — including the HostSync itself, leaving a host with no interfaces, no templates, and no tags.

Root cause

ZabbixSyncBase.try_create() calls api_object().create() without catching the case where the object was created by a concurrent job in the gap between find_by_name() and create().

Fix

Catch the "already exists" error in try_create() and retry via find_by_name() to recover the existing ID:

def try_create(self):
    try:
        result = self.api_object().create(**self.get_create_params())
        return result.get(self.result_key(), [None])[0]
    except Exception as err:
        if 'already exists' in str(err).lower():
            found_by_name = self.find_by_name()
            if len(found_by_name) == 1:
                return found_by_name[0].get(self.id_field.split('.')[-1])
        msg = f'{self.__class__.__name__} creation failed: {err}'
        raise RuntimeError(msg)

Testing

Verified on dev NetBox + Zabbix 7.0.28 with 173 concurrent device/VM sync jobs (HU + JP sites). Before the fix, HostGroupSync failures cascaded to HostSync, leaving hosts with zero interfaces. After the fix, all concurrent creates resolve cleanly — 0 failures from race conditions.

The fix is generic — it applies to all sync classes (HostGroupSync, TemplateSync, ProxySync, etc.) since it lives in the base class.