Files
Varaverk/Tools/zfs_pool_scrub.sh
T

195 lines
7.3 KiB
Bash

#!/bin/bash
# -----------------------------------------------------------------------------------------------
# --------------------------------- ZFS Pool Scrub ---------------------------------------------
# -----------------------------------------------------------------------------------------------
# Triggers a ZFS scrub on all pools (or a specific pool) and waits for completion.
# Sends a notification when scrub completes with a summary of any errors found.
#
# ZFS scrub reads every block on every pool and verifies checksums — it catches
# silent data corruption that would otherwise only surface when you try to read
# the corrupted data. Running monthly is recommended for all ZFS pools.
#
# Usage:
# zfs_pool_scrub.sh — scrub all pools
# zfs_pool_scrub.sh poolname — scrub specific pool only
# zfs_pool_scrub.sh --status — show scrub status for all pools
# zfs_pool_scrub.sh --dry-run — show what would be scrubbed
#
# Pools in ZFS_REPORT_IGNORE_POOLS are skipped unless specified explicitly.
# Scrub runs in background — script polls until complete then reports.
# Safe to run while the pool is in use — scrub does not interrupt normal I/O.
# -----------------------------------------------------------------------------------------------
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "$SCRIPT_DIR/../Master.conf"
source "$SCRIPT_DIR/../common.sh"
parse_args "$@"
TARGET_POOL="${PARSED_ARGS[0]:-}"
# -----------------------------------------------------------------------------------------------
# ━━━ $ICON_GEAR Setup ━━━
# -----------------------------------------------------------------------------------------------
echo ""
echo "━━━ $ICON_GEAR Setup ━━━"
if [[ "$EUID" -ne 0 ]]; then
error "Must be run as root"
exit 1
fi
success "Running as root"
acquire_lock
if ! command -v zpool >/dev/null 2>&1; then
error "ZFS not available on this system"
exit 1
fi
success "ZFS available"
[[ "$DRY_RUN" == true ]] && warn "DRY RUN — no scrubs will be started"
# Build ignore map
declare -A IGNORE_MAP
for pool in "${ZFS_REPORT_IGNORE_POOLS[@]}"; do
[[ -n "$pool" ]] && IGNORE_MAP["$pool"]=1
done
# -----------------------------------------------------------------------------------------------
# ━━━ $ICON_SUMMARY Status ━━━
# -----------------------------------------------------------------------------------------------
if [[ "$SHOW_STATUS" == true ]]; then
echo ""
echo "━━━━━ $ICON_SUMMARY SCRUB STATUS ━━━━━"
zpool list -H -o name 2>/dev/null | while read -r pool; do
SCAN=$(zpool status "$pool" 2>/dev/null | grep "scan:")
echo " $ICON_ZFS $pool$SCAN"
done
echo "━━━━━━━━━━━━━━━━━━━━━━━"
exit 0
fi
# -----------------------------------------------------------------------------------------------
# Build pool list to scrub
# -----------------------------------------------------------------------------------------------
POOLS_TO_SCRUB=()
if [[ -n "$TARGET_POOL" ]]; then
# Specific pool requested — validate it exists
if ! zpool list "$TARGET_POOL" >/dev/null 2>&1; then
error "Pool not found: $TARGET_POOL"
exit 1
fi
POOLS_TO_SCRUB=("$TARGET_POOL")
else
# All pools — skip ignored ones
while IFS= read -r pool; do
[[ -z "$pool" ]] && continue
if [[ -n "${IGNORE_MAP[$pool]:-}" ]]; then
info "Skipping $pool (in ZFS_REPORT_IGNORE_POOLS)"
continue
fi
POOLS_TO_SCRUB+=("$pool")
done < <(zpool list -H -o name 2>/dev/null)
fi
if [[ ${#POOLS_TO_SCRUB[@]} -eq 0 ]]; then
warn "No pools to scrub"
exit 0
fi
info "Pools to scrub: ${POOLS_TO_SCRUB[*]}"
# -----------------------------------------------------------------------------------------------
# ━━━ $ICON_ZFS Start Scrubs ━━━
# -----------------------------------------------------------------------------------------------
echo ""
echo "━━━ $ICON_ZFS Starting ZFS Scrubs ━━━"
START=$(date +%s)
for pool in "${POOLS_TO_SCRUB[@]}"; do
info "$ICON_ZFS Starting scrub on $pool..."
if [[ "$DRY_RUN" == false ]]; then
zpool scrub "$pool" 2>/dev/null && \
success "$pool scrub started" || \
error "Failed to start scrub on $pool"
else
warn "DRY RUN — would scrub: $pool"
fi
done
[[ "$DRY_RUN" == true ]] && {
echo ""
echo "━━━━━ $ICON_SUMMARY ZFS SCRUB SUMMARY ━━━━━"
echo "$ICON_WARN Status: DRY RUN — no scrubs started"
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
exit 0
}
# -----------------------------------------------------------------------------------------------
# ━━━ Poll until complete ━━━
# -----------------------------------------------------------------------------------------------
echo ""
echo "━━━ $ICON_TIME Waiting for scrubs to complete ━━━"
info "Polling every 60 seconds — this may take a while on large pools"
info "Safe to leave running — scrub continues even if this script is stopped"
STILL_RUNNING=true
while [[ "$STILL_RUNNING" == true ]]; do
sleep 60
STILL_RUNNING=false
for pool in "${POOLS_TO_SCRUB[@]}"; do
STATUS=$(zpool status "$pool" 2>/dev/null | grep "scan:" | grep -c "in progress" || true)
if [[ "$STATUS" -gt 0 ]]; then
STILL_RUNNING=true
REPAIRED=$(zpool status "$pool" 2>/dev/null | grep "scan:" | grep -oE "[0-9]+ repaired")
log "$pool — scrub in progress ${REPAIRED:+($REPAIRED)}"
fi
done
done
END=$(date +%s)
success "All scrubs complete"
# -----------------------------------------------------------------------------------------------
# ━━━ $ICON_SUMMARY Results ━━━
# -----------------------------------------------------------------------------------------------
echo ""
echo "━━━ $ICON_ZFS Scrub Results ━━━"
POOLS_OK=()
POOLS_ERRORS=()
for pool in "${POOLS_TO_SCRUB[@]}"; do
SCAN_LINE=$(zpool status "$pool" 2>/dev/null | grep "scan:")
ERRORS=$(zpool status "$pool" 2>/dev/null | grep "errors:" | grep -v "No known data errors")
if [[ -n "$ERRORS" ]]; then
error "$pool$SCAN_LINE"
error "$pool$ERRORS"
POOLS_ERRORS+=("$pool")
else
success "$pool$SCAN_LINE"
POOLS_OK+=("$pool")
fi
done
echo ""
echo "━━━━━ $ICON_SUMMARY ZFS SCRUB SUMMARY ━━━━━"
echo "$ICON_ZFS Pools scrubbed: ${#POOLS_TO_SCRUB[@]}"
echo "$ICON_SUCCESS Clean: ${#POOLS_OK[@]}"
echo "$ICON_ERROR Errors: ${#POOLS_ERRORS[@]}"
echo "$ICON_TIME Duration: $(format_duration $((END - START)))"
echo ""
if [[ ${#POOLS_ERRORS[@]} -gt 0 ]]; then
echo "$ICON_ERROR Status: ERRORS FOUND — ${POOLS_ERRORS[*]}"
notify "ZFS scrub complete on $(hostname) — ERRORS found in pools: ${POOLS_ERRORS[*]}" "ZFS Scrub" "warning"
else
echo "$ICON_DONE Status: $ICON_SUCCESS ALL POOLS CLEAN"
notify "ZFS scrub complete on $(hostname)${#POOLS_OK[@]} pools clean in $(format_duration $((END - START)))" "ZFS Scrub" "normal"
fi
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"