viktor/infra
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fan-control: hold last command through transient HA losses (stop fan flapping)

Browse source 
The actuator dumped the fans to Dell auto on every brief loss of the HA command
(~14% of the time, every few minutes) — crashing them to the ~7100 rpm floor and
bouncing back: the "fans surge then crash then surge" the owner reported. Causes:
the command sensors last_updated going >120s old whenever CPU temp sat flat
(mis-read as stale), plus occasional unavailable blips. Fix: on a missing/stale
command, HOLD the last applied % for up to HA_GRACE_SECS (300s) instead of
falling back, and loosen STALE_SECS 120->1800 (staleness only happens at flat
temp, where the held value is still valid). The 83C CPU CEILING on our own IPMI
read stays the real overheat safety. Verified live: fallback 14% -> 0% over 8h,
command std 16 -> 3, no more rpm floor crashes.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Emil Barzin 2026-06-16 08:07:52 +00:00
parent 082bdfcc77
commit 2cfe338419
1 changed files with 20 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

24
scripts/fan-control.sh
View file

@ -36,7 +36,11 @@ set -uo pipefail
: "${HA_URL:=http://192.168.1.8:8123}" : "${HA_URL:=http://192.168.1.8:8123}"
: "${HA_TOKEN:=}" # long-lived ha-sofia token; empty => Dell auto (no control) : "${HA_TOKEN:=}" # long-lived ha-sofia token; empty => Dell auto (no control)
: "${COMMAND_ENTITY:=sensor.r730_fan_command_pct}" # HA-computed fan %; we only apply it : "${COMMAND_ENTITY:=sensor.r730_fan_command_pct}" # HA-computed fan %; we only apply it
: "${STALE_SECS:=120}" # command older than this => stale => Dell auto : "${STALE_SECS:=1800}" # command older than this => stale. Loose on purpose:
# staleness only happens when CPU temp is flat (so the
# held value is still valid); a rising temp re-renders it.
: "${HA_GRACE_SECS:=300}" # on a transient HA miss, HOLD the last applied % for this
# long before handing the fans to Dell auto (anti-flap)
: "${PUSHGATEWAY_URL:=}" # optional Prometheus Pushgateway base URL : "${PUSHGATEWAY_URL:=}" # optional Prometheus Pushgateway base URL
: "${MAX_IPMI_FAILS:=3}" : "${MAX_IPMI_FAILS:=3}"
: "${MIN_STEP:=3}" # min fan-% change worth an IPMI write (anti-jitter) : "${MIN_STEP:=3}" # min fan-% change worth an IPMI write (anti-jitter)
@ -155,7 +159,7 @@ EOF
main() { main() {
log "fan-control start (actuator; loop=${LOOP_INTERVAL}s ceiling=${CEILING}C cmd=${COMMAND_ENTITY} stale=${STALE_SECS}s dry_run=${DRY_RUN})" log "fan-control start (actuator; loop=${LOOP_INTERVAL}s ceiling=${CEILING}C cmd=${COMMAND_ENTITY} stale=${STALE_SECS}s dry_run=${DRY_RUN})"
trap 'log "exit — restoring Dell auto fan control"; restore_auto' EXIT trap 'log "exit — restoring Dell auto fan control"; restore_auto' EXIT
local current=-1 fails=0 in_fallback=0 cool_since=0 ha_down=0 local current=-1 fails=0 in_fallback=0 cool_since=0 ha_down=0 ha_misses=0
while true; do while true; do
local rpm fan_w; rpm="$(read_fan_rpm)"; rpm="${rpm:-0}"; fan_w="$(fc_fan_watts "$rpm")" local rpm fan_w; rpm="$(read_fan_rpm)"; rpm="${rpm:-0}"; fan_w="$(fc_fan_watts "$rpm")"
@ -190,10 +194,22 @@ main() {
# The setpoint is whatever HA computed. No local math — just apply it. # The setpoint is whatever HA computed. No local math — just apply it.
local cmd; cmd="$(ha_command_pct)" local cmd; cmd="$(ha_command_pct)"
if [[ -z "$cmd" ]]; then if [[ -z "$cmd" ]]; then
(( ha_down == 0 )) && { log "HA command unavailable/stale — Dell auto until it returns"; restore_auto; current=-1; ha_down=1; } ha_misses=$((ha_misses + 1))
push_metrics "$temp" 0 fallback 0 1 "$rpm" "$fan_w" if (( current >= 0 && ha_misses * LOOP_INTERVAL < HA_GRACE_SECS )); then
# Transient HA loss — HOLD the last applied %; do NOT touch the fans. A brief
# command blip (sensor unavailable / stale / fetch hiccup) must not dump the
# fans to Dell auto. The 83C CEILING above (our own IPMI read) is the real
# overheat safety, so holding the last good % is safe.
(( ha_misses == 1 )) && log "HA command miss — holding ${current}% (grace ${HA_GRACE_SECS}s)"
push_metrics "$temp" "$current" applied 0 0 "$rpm" "$fan_w"
else
# Sustained loss (or nothing applied yet) — hand the fans to Dell auto.
(( ha_down == 0 )) && { log "HA command lost (${ha_misses} misses) — Dell auto"; restore_auto; current=-1; ha_down=1; }
push_metrics "$temp" 0 fallback 0 1 "$rpm" "$fan_w"
fi
(( RUN_ONCE == 1 )) && break || { sleep "$LOOP_INTERVAL"; continue; } (( RUN_ONCE == 1 )) && break || { sleep "$LOOP_INTERVAL"; continue; }
fi fi
ha_misses=0
(( ha_down == 1 )) && { log "HA command back (${cmd}%) — resuming"; ha_down=0; } (( ha_down == 1 )) && { log "HA command back (${cmd}%) — resuming"; ha_down=0; }
# Only write when first-run or the change clears MIN_STEP (kills 1-2% jitter). # Only write when first-run or the change clears MIN_STEP (kills 1-2% jitter).