FailedProxy/Varaverk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Replace system free-RAM warning with ARC headroom check

Browse Source 
The memory warning used plain system-wide 'free' RAM (via free -g), a
carryover from before this script was ZFS-specific. ZFS ARC deliberately
consumes most otherwise-unused RAM, so 'free' being low is normal and not
a meaningful signal — it fired a false alarm on 2026-07-05 (free 5.8Gi,
but available a healthy 45Gi).

Replaced ZFS_REPORT_FREE_WARN_GB with ZFS_REPORT_ARC_FREE_WARN_GB: warns
when ARC headroom (ARC_MAX - ARC_CURRENT) drops below threshold, which is
what actually indicates ARC is running out of room to grow. Available RAM
check is unchanged — it's a legitimate system-pressure signal on its own.
This commit is contained in:
Gmer4Lfe
2026-07-11 17:27:59 -04:00
parent 85c3aef1b0
commit ba07634a88
2 changed files with 27 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Monitors/zfs_memory_snapshot.sh
+26 -17
View File
@@ -33,6 +33,14 @@
# available, Docker not responding — those sections skip, the rest still run.
# A partial report is more useful than no report.
#
# ARC Headroom, Not System Free RAM
# ZFS ARC deliberately grows to use most of the RAM the system isn't otherwise
# using — that's the point of a page cache. System-wide "free" RAM being low is
# therefore normal and not a signal of anything, so the memory warning is based
# on ARC headroom (ARC_MAX - ARC_CURRENT) instead — how much room ARC itself has
# left before it hits its configured ceiling. "Available" RAM (which accounts for
# reclaimable cache) is still checked separately as a true system-pressure signal.
#
# ==============================================================================================
# OPERATIONAL MODEL
# ==============================================================================================
@@ -43,10 +51,10 @@
# ZFS_REPORT_IGNORE_POOLS excluded from the report
# (still fully monitored by unRAID — report-only exclusion).
# ARC statistics — current ARC vs max, metadata pressure, hit rate.
# Warns if ARC utilisation exceeds ZFS_REPORT_ARC_WARN_PCT.
# Memory status — total, free, available RAM.
# Warns if free < ZFS_REPORT_FREE_WARN_GB or
# available < ZFS_REPORT_AVAIL_WARN_GB.
# Warns if ARC utilisation exceeds ZFS_REPORT_ARC_WARN_PCT, or if
# ARC headroom (max - current) drops below ZFS_REPORT_ARC_FREE_WARN_GB.
# Memory status — total, free, available RAM (informational only — see note below).
# Warns if available < ZFS_REPORT_AVAIL_WARN_GB.
# Docker memory — top ZFS_REPORT_DOCKER_TOP containers by memory usage.
# Useful for spotting containers approaching watchdog limits.
# Kernel pressure — vmstat snapshot (3 samples).
@@ -102,8 +110,8 @@
# ZFS_REPORT_ARC_WARN_PCT
# Warn if ARC is using more than this percentage of its configured max. (default: 90)
#
# ZFS_REPORT_FREE_WARN_GB
# Warn if free RAM is below this threshold in GB. (default: 10)
# ZFS_REPORT_ARC_FREE_WARN_GB
# Warn if ARC headroom (ARC_MAX - ARC_CURRENT) drops below this many GB. (default: 10)
#
# ZFS_REPORT_AVAIL_WARN_GB
# Warn if available RAM is below this threshold in GB. (default: 20)
@@ -159,7 +167,7 @@ done
log "Identity: $MY_ID ($LOCAL_SERVER_NAME)"
log "Ignoring pools: ${ZFS_REPORT_IGNORE_POOLS[*]:-none}"
log "$ICON_GEAR Config: arc-warn=${ZFS_REPORT_ARC_WARN_PCT}% free-warn=${ZFS_REPORT_FREE_WARN_GB}GB avail-warn=${ZFS_REPORT_AVAIL_WARN_GB}GB docker-top=${ZFS_REPORT_DOCKER_TOP}"
log "$ICON_GEAR Config: arc-warn=${ZFS_REPORT_ARC_WARN_PCT}% arc-free-warn=${ZFS_REPORT_ARC_FREE_WARN_GB}GB avail-warn=${ZFS_REPORT_AVAIL_WARN_GB}GB docker-top=${ZFS_REPORT_DOCKER_TOP}"
# Tee output to log file unless dry run
if [[ "$DRY_RUN" == false ]]; then
@@ -178,7 +186,7 @@ if [[ "$SHOW_STATUS" == true ]]; then
echo "$ICON_HOST Identity: $MY_ID ($LOCAL_SERVER_NAME)"
echo "$ICON_ZFS Log file: $ZFS_REPORT_LOG"
echo "$ICON_ZFS ARC warn: ${ZFS_REPORT_ARC_WARN_PCT}%"
echo "$ICON_MEM Free RAM warn: ${ZFS_REPORT_FREE_WARN_GB}GB"
echo "$ICON_ZFS ARC free warn: ${ZFS_REPORT_ARC_FREE_WARN_GB}GB"
echo "$ICON_MEM Avail warn: ${ZFS_REPORT_AVAIL_WARN_GB}GB"
echo "$ICON_CONTAINERS Docker top: $ZFS_REPORT_DOCKER_TOP"
echo "$ICON_ZFS Ignore pools: ${ZFS_REPORT_IGNORE_POOLS[*]:-none}"
@@ -266,11 +274,13 @@ else
ARC_META_GB=$(bytes_to_gb "$ARC_META_USED" 1)
ARC_PCT=$(awk "BEGIN {printf \"%.1f\", $ARC_SIZE * 100 / $ARC_MAX}")
ARC_PCT_INT=$(printf "%.0f" "$ARC_PCT")
ARC_FREE_GB=$(awk "BEGIN {printf \"%d\", ($ARC_MAX - $ARC_SIZE) / 1073741824}")
echo " $ICON_ZFS ARC Max: ${ARC_MAX_GB}GB"
echo " $ICON_ZFS ARC Current: ${ARC_CUR_GB}GB"
echo " $ICON_ZFS ARC Meta Used: ${ARC_META_GB}GB"
echo " $ICON_ZFS ARC Utilization: ${ARC_PCT}%"
echo " $ICON_ZFS ARC Free: ${ARC_FREE_GB}GB"
if [[ "$ARC_PCT_INT" -ge "$ZFS_REPORT_ARC_WARN_PCT" ]]; then
warn "ARC utilization ${ARC_PCT}% — above ${ZFS_REPORT_ARC_WARN_PCT}% threshold"
@@ -279,6 +289,13 @@ else
log "ARC utilization ${ARC_PCT}% — within threshold ✅"
fi
if [[ "$ARC_FREE_GB" -lt "$ZFS_REPORT_ARC_FREE_WARN_GB" ]]; then
warn "ARC free ${ARC_FREE_GB}GB — below ${ZFS_REPORT_ARC_FREE_WARN_GB}GB threshold"
WARNINGS+=("ARC headroom low: ${ARC_FREE_GB}GB")
else
log "ARC free ${ARC_FREE_GB}GB — within threshold ✅"
fi
echo ""
META_MRU_GHOST=$(awk '/^mru_ghost_metadata / {print $3}' \
/proc/spl/kstat/zfs/arcstats 2>/dev/null || echo 0)
@@ -302,20 +319,12 @@ echo "━━━ $ICON_MEM Memory Status ━━━"
FREE_HUMAN=$(free -h | awk '/Mem:/ {print $4}')
AVAIL_HUMAN=$(free -h | awk '/Mem:/ {print $7}')
TOTAL_HUMAN=$(free -h | awk '/Mem:/ {print $2}')
FREE_GB=$(free -g | awk '/Mem:/ {print $4}')
AVAIL_GB=$(free -g | awk '/Mem:/ {print $7}')
echo " $ICON_MEM Total RAM: $TOTAL_HUMAN"
echo " $ICON_MEM Free RAM: $FREE_HUMAN"
echo " $ICON_MEM Free RAM: $FREE_HUMAN (informational — ARC intentionally uses most of this)"
echo " $ICON_MEM Available RAM: $AVAIL_HUMAN"
if [[ "$FREE_GB" -lt "$ZFS_REPORT_FREE_WARN_GB" ]]; then
warn "Free RAM ${FREE_HUMAN} — below ${ZFS_REPORT_FREE_WARN_GB}GB threshold"
WARNINGS+=("Low free RAM: ${FREE_HUMAN}")
else
log "Free RAM ${FREE_HUMAN} — within threshold ✅"
fi
if [[ "$AVAIL_GB" -lt "$ZFS_REPORT_AVAIL_WARN_GB" ]]; then
warn "Available RAM ${AVAIL_HUMAN} — below ${ZFS_REPORT_AVAIL_WARN_GB}GB threshold"
WARNINGS+=("Low available RAM: ${AVAIL_HUMAN}")