Vpp fall back to load first?

[ridax67]

Are Vpp control Are Vpp control set every 15 minutes? It has not been deemed unnecessary if it is already in the correct position? It seems the 20 minutes Vpp timeout kicks in (5 minutes after an update is expected).

Did you implement some generic modbus register write that tried to stop writes when unnecessary, or do I recall wrongly? The vpp control register needs to be updated every 15 minute, even if it is already set, unless you want to be in load first.

Could be my fault also, if something crashes or something, but I can’t find any reason here.

bess-debug-2026-07-27-075654.md

[ridax67]

[johanzander]

Root cause found

You're on solax_modbus in VPP mode with a 20-minute inverter fallback timer (vpp_time). The bug is in core/bess/solax_modbus_growatt_controller.py, _apply_period_vpp():

command_changed = (
    remote_control_enabled != self._last_written_vpp_remote_control
    or (remote_control_enabled and power_pct != self._last_written_vpp_power)
)
if not command_changed:
    return True, ""

This skips the write to the inverter (and with it, the only call that rewrites the fallback timer register) whenever the VPP command hasn't changed since last period. During a stable run of identical periods — e.g. LOAD_SUPPORT unchanged for 2+ hours, as seen in your debug bundle — the timer never gets refreshed. After 20 minutes with no rewrite, the inverter's own dead-man's-switch lapses and it reverts to load_first, which is exactly what you're seeing.

This has been present since VPP mode was introduced (#297) — it's not a regression from a recent release, and it's unrelated to #402 (that dedup only affects the TOU/Growatt-cloud write path). The module's own docstring even documents the intended behavior ("the fallback timer is rewritten every active period") — the implementation just never matched it.

Fix: keep writing every period while a VPP command is actively enabled (refreshing the timer), and only skip the write when there's truly nothing active (load_first, unchanged). This preserves the write-minimization for the idle case while eliminating the timeout during active control.

I have a fix ready and tests updated/passing locally; will follow up with a PR.

[ridax67]

The fix works