245 lines
12 KiB
Bash
245 lines
12 KiB
Bash
#!/bin/bash
|
|
# -----------------------------------------------------------------------------------------------
|
|
# --------------------------------- Health Digest ----------------------------------------------
|
|
# -----------------------------------------------------------------------------------------------
|
|
# Aggregates system health data from across the ecosystem into a single digest report.
|
|
# Reads existing state files — no new writes to flash drive.
|
|
#
|
|
# Three profiles controlled by DIGEST_PROFILE in Master.conf:
|
|
# always — sends every run regardless of findings (schedule daily for daily digest)
|
|
# smart — sends only if something worth reporting was found (intelligent filtering)
|
|
# weekly — sends once per week on DIGEST_DAY regardless of schedule frequency
|
|
#
|
|
# The cron schedule stays the same regardless of profile — just change DIGEST_PROFILE
|
|
# in Master.conf to switch behavior. Run daily, profile controls when it actually notifies.
|
|
#
|
|
# Data sources (reads only — no writes):
|
|
# /tmp/transcode_state.db — ramdisk symlink and usage
|
|
# /tmp/container_watchdog_state.db — active container strikes
|
|
# /tmp/system_watchdog_state.db — active system strikes
|
|
# /boot/config/failover_state.db — current failover state
|
|
# /boot/config/system_watchdog_failed.db — container skip list
|
|
# /boot/config/bandwidth_history.db — recent transfer totals
|
|
# SSL certs via openssl (live check) — days remaining per domain
|
|
#
|
|
# All configuration in Master.conf under Health Digest section.
|
|
# Supports --dry-run to generate report without sending notification.
|
|
# -----------------------------------------------------------------------------------------------
|
|
|
|
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
|
|
|
source "$SCRIPT_DIR/../Master.conf"
|
|
source "$SCRIPT_DIR/../common.sh"
|
|
|
|
parse_args "$@"
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# ━━━ $ICON_GEAR Setup ━━━
|
|
# -----------------------------------------------------------------------------------------------
|
|
echo ""
|
|
echo "━━━ $ICON_GEAR Setup ━━━"
|
|
|
|
success "Health Digest — profile: $DIGEST_PROFILE"
|
|
[[ "$DRY_RUN" == true ]] && warn "DRY RUN — report generated but no notification sent"
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# Profile check — should we send today?
|
|
# -----------------------------------------------------------------------------------------------
|
|
SHOULD_SEND=false
|
|
|
|
case "$DIGEST_PROFILE" in
|
|
always)
|
|
SHOULD_SEND=true
|
|
log "Profile: always — will send"
|
|
;;
|
|
weekly)
|
|
TODAY_NAME=$(date '+%A')
|
|
if [[ "$TODAY_NAME" == "$DIGEST_DAY" ]]; then
|
|
SHOULD_SEND=true
|
|
log "Profile: weekly — today is $DIGEST_DAY, will send"
|
|
else
|
|
info "Profile: weekly — today is $TODAY_NAME, digest day is $DIGEST_DAY — skipping"
|
|
exit 0
|
|
fi
|
|
;;
|
|
smart)
|
|
log "Profile: smart — will evaluate findings before deciding"
|
|
SHOULD_SEND=false # determined after gathering data
|
|
;;
|
|
*)
|
|
warn "Unknown DIGEST_PROFILE: $DIGEST_PROFILE — defaulting to weekly behavior"
|
|
TODAY_NAME=$(date '+%A')
|
|
[[ "$TODAY_NAME" == "$DIGEST_DAY" ]] && SHOULD_SEND=true
|
|
;;
|
|
esac
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# DATA GATHERING
|
|
# -----------------------------------------------------------------------------------------------
|
|
echo ""
|
|
echo "━━━ $ICON_DIGEST Gathering System Data ━━━"
|
|
|
|
FINDINGS=() # things worth noting
|
|
ISSUES=() # things that need attention
|
|
DIGEST_LINES=() # full report lines
|
|
|
|
# ── Failover State ──────────────────────────────────────────────────────────────────────────
|
|
if [[ -f "$FAILOVER_STATE_FILE" ]]; then
|
|
FAILOVER_STATE=$(grep "^state=" "$FAILOVER_STATE_FILE" 2>/dev/null | cut -d= -f2)
|
|
FAILOVER_CHANGE=$(grep "^last_change=" "$FAILOVER_STATE_FILE" 2>/dev/null | cut -d= -f2)
|
|
DIGEST_LINES+=("$ICON_FAILOVER Failover: $FAILOVER_STATE (last change: ${FAILOVER_CHANGE:-unknown})")
|
|
|
|
if [[ "$FAILOVER_STATE" != "NORMAL" && -n "$FAILOVER_STATE" ]]; then
|
|
ISSUES+=("Failover state: $FAILOVER_STATE")
|
|
[[ "$DIGEST_SMART_ON_FAILOVER" == true ]] && SHOULD_SEND=true
|
|
fi
|
|
else
|
|
DIGEST_LINES+=("$ICON_FAILOVER Failover: state file not found")
|
|
fi
|
|
|
|
# ── Transcode Ramdisk ───────────────────────────────────────────────────────────────────────
|
|
if mountpoint -q "$RAMDISK_PATH" 2>/dev/null; then
|
|
RAMDISK_USED_KB=$(df "$RAMDISK_PATH" --output=used | tail -1 | tr -d ' ')
|
|
RAMDISK_USED_GB=$(awk "BEGIN {printf \"%.2f\", $RAMDISK_USED_KB / 1048576}")
|
|
RAMDISK_AVAIL_KB=$(df "$RAMDISK_PATH" --output=avail | tail -1 | tr -d ' ')
|
|
RAMDISK_AVAIL_GB=$(awk "BEGIN {printf \"%.2f\", $RAMDISK_AVAIL_KB / 1048576}")
|
|
SYMLINK_TARGET=$(readlink "$TRANSCODE_LINK" 2>/dev/null || echo "unknown")
|
|
DIGEST_LINES+=("$ICON_RAM Transcodes: ${RAMDISK_USED_GB}GB used / ${RAMDISK_AVAIL_GB}GB free → $SYMLINK_TARGET")
|
|
else
|
|
DIGEST_LINES+=("$ICON_RAM Transcodes: ramdisk not mounted")
|
|
ISSUES+=("Ramdisk not mounted")
|
|
fi
|
|
|
|
# ── Container Watchdog Strikes ──────────────────────────────────────────────────────────────
|
|
if [[ -f "$WATCHDOG_STATE_FILE" ]]; then
|
|
ACTIVE_STRIKES=$(grep -v ":0$" "$WATCHDOG_STATE_FILE" 2>/dev/null | wc -l)
|
|
if [[ "$ACTIVE_STRIKES" -gt 0 ]]; then
|
|
STRIKE_LIST=$(grep -v ":0$" "$WATCHDOG_STATE_FILE" 2>/dev/null | tr '\n' ' ')
|
|
DIGEST_LINES+=("$ICON_WATCHDOG Container strikes: $ACTIVE_STRIKES active — $STRIKE_LIST")
|
|
FINDINGS+=("Container watchdog: $ACTIVE_STRIKES active strikes")
|
|
[[ "$DIGEST_SMART_ON_WATCHDOG" == true ]] && SHOULD_SEND=true
|
|
else
|
|
DIGEST_LINES+=("$ICON_WATCHDOG Container watchdog: no active strikes")
|
|
fi
|
|
fi
|
|
|
|
# ── System Watchdog Strikes ─────────────────────────────────────────────────────────────────
|
|
if [[ -f "$SYS_WATCHDOG_STATE_FILE" ]]; then
|
|
SYS_STRIKES=$(grep -v ":0$" "$SYS_WATCHDOG_STATE_FILE" 2>/dev/null | wc -l)
|
|
if [[ "$SYS_STRIKES" -gt 0 ]]; then
|
|
SYS_STRIKE_LIST=$(grep -v ":0$" "$SYS_WATCHDOG_STATE_FILE" 2>/dev/null | tr '\n' ' ')
|
|
DIGEST_LINES+=("$ICON_REBOOT_SMART System strikes: $SYS_STRIKES active — $SYS_STRIKE_LIST")
|
|
FINDINGS+=("System watchdog: $SYS_STRIKES active strikes")
|
|
[[ "$DIGEST_SMART_ON_WATCHDOG" == true ]] && SHOULD_SEND=true
|
|
else
|
|
DIGEST_LINES+=("$ICON_REBOOT_SMART System watchdog: no active strikes")
|
|
fi
|
|
fi
|
|
|
|
# ── Container Skip List ─────────────────────────────────────────────────────────────────────
|
|
if [[ -f "$SYS_WATCHDOG_FAILED_FILE" ]] && [[ -s "$SYS_WATCHDOG_FAILED_FILE" ]]; then
|
|
SKIP_COUNT=$(wc -l < "$SYS_WATCHDOG_FAILED_FILE")
|
|
SKIP_LIST=$(cat "$SYS_WATCHDOG_FAILED_FILE" | tr '\n' ' ')
|
|
DIGEST_LINES+=("$ICON_NOT_RUNNING Skip list: $SKIP_COUNT containers — $SKIP_LIST")
|
|
ISSUES+=("Containers on skip list: $SKIP_LIST")
|
|
SHOULD_SEND=true
|
|
else
|
|
DIGEST_LINES+=("$ICON_RUNNING Skip list: empty — all containers healthy")
|
|
fi
|
|
|
|
# ── Bandwidth (yesterday's total) ───────────────────────────────────────────────────────────
|
|
if [[ -f "$BANDWIDTH_LOG" ]] && [[ -s "$BANDWIDTH_LOG" ]]; then
|
|
YESTERDAY=$(date -d "yesterday" '+%Y-%m-%d')
|
|
YESTERDAY_BYTES=$(awk -F'|' -v d="$YESTERDAY" '$1==d{sum+=$3} END{print sum+0}' "$BANDWIDTH_LOG")
|
|
YESTERDAY_GB=$(awk "BEGIN {printf \"%.2f\", $YESTERDAY_BYTES / 1073741824}")
|
|
OVER_WARN=$(awk "BEGIN {print ($YESTERDAY_BYTES > $BANDWIDTH_WARN_GB * 1073741824) ? 1 : 0}")
|
|
|
|
if [[ "$OVER_WARN" == "1" ]]; then
|
|
DIGEST_LINES+=("$ICON_BANDWIDTH Yesterday's transfers: ${YESTERDAY_GB}GB ← exceeded ${BANDWIDTH_WARN_GB}GB threshold")
|
|
FINDINGS+=("High bandwidth day: ${YESTERDAY_GB}GB transferred")
|
|
[[ "$DIGEST_SMART_ON_BANDWIDTH" == true ]] && SHOULD_SEND=true
|
|
else
|
|
DIGEST_LINES+=("$ICON_BANDWIDTH Yesterday's transfers: ${YESTERDAY_GB}GB")
|
|
fi
|
|
else
|
|
DIGEST_LINES+=("$ICON_BANDWIDTH Bandwidth: no data yet")
|
|
fi
|
|
|
|
# ── SSL Certificates ────────────────────────────────────────────────────────────────────────
|
|
if [[ ${#CERT_MONITOR_DOMAINS[@]} -gt 0 ]] && command -v openssl >/dev/null 2>&1; then
|
|
CERT_ISSUES=()
|
|
for domain in "${CERT_MONITOR_DOMAINS[@]}"; do
|
|
[[ -z "$domain" ]] && continue
|
|
expiry_str=$(echo | timeout "$CERT_TIMEOUT" openssl s_client \
|
|
-connect "${domain}:443" -servername "$domain" \
|
|
2>/dev/null | openssl x509 -noout -enddate 2>/dev/null | cut -d= -f2)
|
|
|
|
if [[ -n "$expiry_str" ]]; then
|
|
expiry_epoch=$(date -d "$expiry_str" +%s 2>/dev/null)
|
|
days_remaining=$(( (expiry_epoch - $(date +%s)) / 86400 ))
|
|
|
|
if [[ "$days_remaining" -le "$CERT_CRIT_DAYS" ]]; then
|
|
CERT_ISSUES+=("$domain: ${days_remaining}d CRITICAL")
|
|
SHOULD_SEND=true
|
|
elif [[ "$days_remaining" -le "$CERT_WARN_DAYS" ]]; then
|
|
CERT_ISSUES+=("$domain: ${days_remaining}d WARNING")
|
|
[[ "$DIGEST_SMART_ON_CERT_WARN" == true ]] && SHOULD_SEND=true
|
|
fi
|
|
fi
|
|
done
|
|
|
|
if [[ ${#CERT_ISSUES[@]} -gt 0 ]]; then
|
|
DIGEST_LINES+=("$ICON_CERT Certificates: ${CERT_ISSUES[*]}")
|
|
FINDINGS+=("Cert issues: ${CERT_ISSUES[*]}")
|
|
else
|
|
DIGEST_LINES+=("$ICON_CERT Certificates: all healthy")
|
|
fi
|
|
fi
|
|
|
|
# ── Smart profile final decision ────────────────────────────────────────────────────────────
|
|
if [[ "$DIGEST_PROFILE" == "smart" && "$SHOULD_SEND" == false ]]; then
|
|
info "Profile: smart — no findings worth reporting — skipping notification"
|
|
echo ""
|
|
echo "━━━━━ $ICON_SUMMARY DIGEST SUMMARY ━━━━━"
|
|
echo "$ICON_SUCCESS Everything looks healthy — no digest sent (smart profile)"
|
|
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
|
|
exit 0
|
|
fi
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# ━━━ $ICON_DIGEST Build and Send Digest ━━━
|
|
# -----------------------------------------------------------------------------------------------
|
|
echo ""
|
|
echo "━━━ $ICON_DIGEST Health Digest ━━━"
|
|
|
|
DIGEST_DATE=$(date '+%Y-%m-%d %H:%M:%S')
|
|
DIGEST_HOST=$(hostname)
|
|
|
|
# Build notification message
|
|
NOTIFY_MSG="Health Digest — $DIGEST_HOST — $DIGEST_DATE"
|
|
if [[ ${#ISSUES[@]} -gt 0 ]]; then
|
|
NOTIFY_MSG+=" | Issues: ${ISSUES[*]}"
|
|
fi
|
|
if [[ ${#FINDINGS[@]} -gt 0 ]]; then
|
|
NOTIFY_MSG+=" | Findings: ${FINDINGS[*]}"
|
|
fi
|
|
|
|
# Print full digest to console
|
|
for line in "${DIGEST_LINES[@]}"; do
|
|
echo " $line"
|
|
done
|
|
|
|
echo ""
|
|
echo "━━━━━ $ICON_SUMMARY DIGEST SUMMARY ━━━━━"
|
|
echo "$ICON_DIGEST Profile: $DIGEST_PROFILE"
|
|
echo "$ICON_TIME Generated: $DIGEST_DATE"
|
|
echo "$ICON_ERROR Issues: ${#ISSUES[@]}"
|
|
echo "$ICON_WARN Findings: ${#FINDINGS[@]}"
|
|
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
|
|
|
|
if [[ "$DRY_RUN" == true ]]; then
|
|
warn "DRY RUN — digest generated but not sent"
|
|
elif [[ "$SHOULD_SEND" == true ]]; then
|
|
notify "$NOTIFY_MSG" "Health Digest" "$([[ ${#ISSUES[@]} -gt 0 ]] && echo "warning" || echo "normal")"
|
|
success "Digest sent"
|
|
fi |