[pve-devel] [PATCH ha-manager 2/2] add api getter/setter for node maintenance mode

[Thomas Lamprecht]

On 26/08/2025 18:02, Thomas Skinner wrote:
>> The more I think about it, I'm somewhat inclined to register the whole
>> subclass under /nodes/{node}/ha/... instead of /cluster/ha/nodes/... to
>> not duplicate the information of a nodelist, which is provided by
>> pve-cluster and there would be something wrong with the HA Manager
>> anyway if those are out-of-sync and all the HA-related information can
>> then live in a node-specific "ha" subdirectory.
>>
>> Then this API handler wouldn't be needed anymore at all.
> If this is how Proxmox would like it, that makes it much easier IMO. I
> wasn't sure which one to include it under, so I put it under the
> cluster path since a cluster must be created to even have the HA
> featureset. I'm good with either, I just need to know so that I know
> what to update for a v2 series.

We might be already all on the same page, so just to be sure and a
bit more context to avoid making this to HA specific:

Reporting the current (maintenance) status should go into either (or
both of) the /cluster/ha/status and/or the /cluster/resources API
endpoint.

As there are some light-weight plans to provide maintenance mode
for non-ha services, I'd indeed place that outside of a ha specific
path now already, might be even fine to use a dedicated
/nodes/{node}/maintenance API endpoint. For now the description should
state that it's only affecting HA, the non-HA parts are probably not
very complicated work, but got quite a few edge cases and design
decisions that are better done in a separate series building on top
of such API infrastructure you add here.