Re: [pve-devel] [PATCH v2 guest-common/manager 0/4] vzdump: add 'notification-mode' parameter

[Fiona Ebner <f.ebner@proxmox.com>]

Am 21.11.23 um 13:13 schrieb Lukas Wagner:
> On 11/21/23 11:55, Fiona Ebner wrote:
>> Am 21.11.23 um 11:22 schrieb Lukas Wagner:
>>> This should provide a fix/workaround for the users' reports of
>>>    - double notifications (these happened in case mailto was set to
>>> the same address
>>>      as root@pam)
>>
>> Can't we detect and avoid this more easily?
> 
> There would be other ways to solve this, yes. I could also deduplicate
> email-addresses in the backend - which isn't is as trivial as it sounds,
> since the 'legacy' mails are sent via separate, temporary target of
> type 'sendmail' - so essentially I'd need to have 'cross-target' context
> or something alike.
> 

Okay, I hoped you could get the information about whether the 'mailto'
address is already notified via existing targets more easily.

>>
>>>    - notifications always being sent, even if 'mailnotification' is
>>> set to failure
>>
>> Can't we just treat 'failure' mode as always defaulting to legacy
>> sendmail? And properly deprecate the setting, showing a warning/info
>> that new notification system is not used if set to 'failure' for both
>> CLI and UI. And maybe not even allow setting it for new jobs/manual
>> backups?
> 
> I think an explicit switch here is much more obvious and predictable to
> the user.
> 

But also more complexity/mental load for both them and us. And having a
deprecation warning would also be obvious and predictable ;) Sure, we
(most likely) can get rid of the extra switch with PVE 9, was just
hoping for something simpler. But if that's not easily possible, then
let's go with the switch.

> Personally I think we should wait a bit before deprecating/disallowing
> 'mailto' for new backup jobs. This gives us time to polish the UX of
> creating matchers etc. without 'forcing' the user into the new system.
> 

I didn't mean deprecating 'mailto' like that, just 'mailnotification'
(being set to 'failure').