Re: [PATCH-SERIES ha-manager 0/7] improve handling of maintenance nodes

["Daniel Kral" <d.kral@proxmox.com>]

Superseded-by: https://lore.proxmox.com/pve-devel/20261106095650.72865-1-d.kral@proxmox.com/

On Wed Apr 22, 2026 at 12:00 PM CEST, Daniel Kral wrote:
> As reported by a recent Proxmox forum post [0], there are some cases
> where either HA resources are not moved away from maintenance nodes or
> are not moved back to the maintenance nodes after these are put out of
> maintenance again.
>
> Even though we cannot resolve all situations (for example, the affinity
> rules constrain the HA resource so that it cannot be moved anywhere but
> the maintenance node), this patch series improves the handling with:
>
> - log warnings if HA resources cannot be moved to a replacement node
> - make HA resources with fallback enabled move back to their previous
>   maintenance node
> - make HA resource bundles move back to their previous maintenance node
> - try all available, effective priority classes for an HA resource while
>   applying the negative resource affinity rules
>
> The last change makes the system more consistent overall, but might
> introduce some unintended node placements in highly constrained
> scenarios because of how the HA Manager currently resolves these node
> placements individually per-HA resource. This should be improved upon in
> a future patch series (this bugzilla entry [1] might also be relevant).
>
> [0] https://forum.proxmox.com/threads/182890/
> [1] https://bugzilla.proxmox.com/show_bug.cgi?id=7475