Re: [PATCH manager v1 1/1] api: startall: print info message if guest is skipped due to no onboot

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

Am 02.03.26 um 4:16 PM schrieb Michael Köppl:
> On Mon Mar 2, 2026 at 4:10 PM CET, Fiona Ebner wrote:
>> Am 02.03.26 um 4:02 PM schrieb Michael Köppl:
>>> On Mon Mar 2, 2026 at 3:09 PM CET, Fiona Ebner wrote:
>>>> Am 02.03.26 um 2:49 PM schrieb Michael Köppl:
>>>>> The documentation states that startall only starts guests with
>>>>> onboot=1 by default, and that this behavior can be overridden using the
>>>>> force parameter. However, when startall is invoked via the pvenode CLI
>>>>> without the force parameter, the Bulk Start task silently completes with
>>>>> just "TASK OK", giving no indication of why certain VMs were not started.
>>>>> The added informational message addresses this by clearly communicating
>>>>> to users why those VMs were skipped.
>>>>>
>>>>> Signed-off-by: Michael Köppl <m.koeppl@proxmox.com>
>>>>> ---
>>>>> I encountered this while using startall and stopall myself and while
>>>>> RTFM would indeed have helped, I still felt that an informational
>>>>> message would improve the user's experience, especially since stopall
>>>>> will stop all VMs without force=1, whereas startall requires the force
>>>>> param. I only added the informational messages and did not change any
>>>>> behavior because the behavior makes sense to me after thinking about
>>>>> it some more.
>>>>>
>>>>>  PVE/API2/Nodes.pm | 7 ++++++-
>>>>>  1 file changed, 6 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/PVE/API2/Nodes.pm b/PVE/API2/Nodes.pm
>>>>> index 5bd6fe492..3faa1e800 100644
>>>>> --- a/PVE/API2/Nodes.pm
>>>>> +++ b/PVE/API2/Nodes.pm
>>>>> @@ -1969,7 +1969,12 @@ sub get_start_stop_list {
>>>>>      my $resList = {};
>>>>>      foreach my $vmid (keys %$vmlist) {
>>>>>          my $conf = $vmlist->{$vmid}->{conf};
>>>>> -        next if $autostart && !$conf->{onboot};
>>>>> +
>>>>> +        if ($autostart && !$conf->{onboot}) {
>>>>> +            print
>>>>> +                "skipping $vmid because 'onboot' is not set in guest config, use 'force' parameter to override\n";
>>>>> +            next;
>>>>> +        }
>>>>
>>>> I think printing it for every single guest without onboot is too much,
>>>> because there could be thousands of such guests. One message at the
>>>> beginning of the API call should be enough.
>>>>
>>>
>>> Yeah, I wasn't entirely sure printing it for every guest is a good idea
>>> either. Thanks for the feedback. I guess something like "skipping guests
>>> without 'onboot' set in guest config, use 'force' param to override"
>>> once at the beginning?
>>>
>>>> And I feel like the invocation from pve-guests.service should not have
>>>> such a message end up in syslog to avoid confusion. It uses
>>>> /usr/bin/pvesh --nooutput create /nodes/localhost/startall
>>>> so maybe this is already the case. Could you check?
>>>>
>>>
>>> I agree, but --nooutput does not seem to prevent this. I'll have a look
>>> how this can be avoided.
>>
>> If it can't easily be avoided, I guess the message is best formulated in
>> a purely descriptive way, i.e. without "use to override", and rather
>> just mention that it's because force is not set.
> 
> Ack, thanks! I'll send a v2.

If you only want it for CLI you can check the rpcenv type. I guess we
can expect API users to read the description before using the endpoint.
In the UI we always use it anyways (from a quick glance).