[pve-devel] [PATCH pve-esxi-import-tools] fix #6347: do not break vm listing on invalid configs

[Daniel Kral]

On 4/23/25 16:04, Daniel Herzig wrote:
> Daniel Kral <d.kral at proxmox.com> writes:
> 
>> On 4/23/25 15:03, Daniel Herzig wrote:
>>> Move the queries for vCLS machines and empty datastore strings into
>>> the already existing try-except routine to save `listvms.py` from
>>> breaking when `vm.config` is `None`.
>>> Do not store the result of the `any` function in a variable anymore,
>>> do the same for the datastore query to shorten code.
>>> Prepare more verbose logging.
>>> Signed-off-by: Daniel Herzig <d.herzig at proxmox.com>
>>
>> Oops, I should have prefixed my patch with a "fix" at least, but I've
>> sent two patches some days ago for this ;):
>>
>> https://lore.proxmox.com/pve-devel/20250411150634.253684-1-d.kral@proxmox.com/T/#u
>>
> Ha, I missed that one, you were too quick ;)
> 
> It turned out that the problem was on the specific ESXI-side config
> (`vm.config` is there on ESXI-6.7.x as well, without it we could not do
> anything anyway -- `fetch_and_update_vm_data` calls it more than
> once inherently). But it's `None` for sure if there's an invalid vm config on the ESXI side.

Thanks for the info!

I was wondering why it didn't work for them (in the forum) and have only 
figured that there was something undocumented in the PyVim API about 
when the OptionType is None, but with either patch it should be fine now :).

> 
> OP on bugzilla reported that once these invalid VMs were removed from
> ESXI, 'standard 0.7.3' started working as expected.
> 
> But for the earlier vCLS commit I totally missed that
> possibility (shame on me) and placed the queries 'unprotected' (unlike
> `fetch_and_update_vm_data`).
> 
>> But it's also good to just extend the scope of the try-catch here
>> without the additional check for whether the vm.config has a value or
>> not.