305 lines
12 KiB
Bash
305 lines
12 KiB
Bash
#!/bin/bash
|
|
# -----------------------------------------------------------------------------------------------
|
|
# --------------------------------- Docker Watchdog --------------------------------------------
|
|
# -----------------------------------------------------------------------------------------------
|
|
# Self-healing watchdog for Docker containers — monitors memory, CPU and HTTP responsiveness.
|
|
# Restarts containers that exceed configured thresholds using a strike system for CPU and
|
|
# responsiveness checks to avoid restarting on brief spikes.
|
|
#
|
|
# Behaviour:
|
|
# Memory — immediate restart if hard limit is exceeded
|
|
# CPU — strike system, restarts after CPU_FAIL_LIMIT consecutive over-threshold checks
|
|
# HTTP — strike system, restarts after RESP_FAIL_LIMIT consecutive failed curl checks
|
|
#
|
|
# Strike system:
|
|
# Strikes persist between runs via WATCHDOG_STATE_FILE (/tmp — resets on reboot)
|
|
# Strike cadence depends on cron schedule:
|
|
# Every 15min + 2 strikes = 30min sustained abuse before restart
|
|
# Every 10min + 2 strikes = 20min sustained abuse before restart
|
|
# Every 5min + 2 strikes = 10min sustained abuse before restart
|
|
#
|
|
# All configuration lives in Master.conf under the Docker Watchdog section.
|
|
# Supports --dry-run to show what would be restarted without taking any action.
|
|
# -----------------------------------------------------------------------------------------------
|
|
|
|
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
|
|
|
source "$SCRIPT_DIR/../Master.conf"
|
|
source "$SCRIPT_DIR/../common.sh"
|
|
|
|
parse_args "$@"
|
|
|
|
# Auto-detect total CPU cores for normalisation
|
|
TOTAL_CORES=$(nproc)
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# ━━━ $ICON_GEAR Setup ━━━
|
|
# -----------------------------------------------------------------------------------------------
|
|
echo ""
|
|
echo "━━━ $ICON_GEAR Setup ━━━"
|
|
|
|
# ROOT CHECK
|
|
if [[ "$EUID" -ne 0 ]]; then
|
|
error "Must be run as root"
|
|
exit 1
|
|
fi
|
|
|
|
success "Running as root"
|
|
info "$ICON_WATCHDOG Watchdog initialising — $TOTAL_CORES cores detected"
|
|
|
|
# Ensure state file exists
|
|
touch "$WATCHDOG_STATE_FILE" 2>/dev/null || {
|
|
error "Cannot create state file: $WATCHDOG_STATE_FILE"
|
|
exit 1
|
|
}
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# ━━━ $ICON_SUMMARY Status ━━━
|
|
# -----------------------------------------------------------------------------------------------
|
|
if [[ "$SHOW_STATUS" == true ]]; then
|
|
echo ""
|
|
echo "━━━━━ $ICON_SUMMARY STATUS ━━━━━"
|
|
echo "$ICON_WATCHDOG Containers monitored: ${!WATCHDOG_CONTAINERS[*]}"
|
|
echo "$ICON_MEM Soft mem threshold: ${SOFT_MEM_THRESHOLD}% of per-container limit"
|
|
echo "$ICON_ZFS CPU soft threshold: ${SOFT_CPU_THRESHOLD}%"
|
|
echo "$ICON_ZFS CPU hard threshold: ${HARD_CPU_THRESHOLD}%"
|
|
echo "$ICON_RETRY CPU fail limit: ${CPU_FAIL_LIMIT} strikes"
|
|
echo "$ICON_PING Resp fail limit: ${RESP_FAIL_LIMIT} strikes"
|
|
echo "$ICON_TIME Curl timeout: ${CURL_TIMEOUT}s"
|
|
echo "$ICON_GEAR Dry Run: $DRY_RUN"
|
|
echo "━━━━━━━━━━━━━━━━━━━━━━━"
|
|
exit 0
|
|
fi
|
|
|
|
[[ "$DRY_RUN" == true ]] && warn "DRY RUN — no containers will be restarted"
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# STATE HELPERS
|
|
# Reads and writes per-container strike counts to the state file.
|
|
# State file format: container:metric:count
|
|
# -----------------------------------------------------------------------------------------------
|
|
|
|
# Returns current strike count for a container/metric pair.
|
|
# Usage: get_strikes "Emby" "CPU"
|
|
get_strikes() {
|
|
local container="$1" metric="$2"
|
|
grep -E "^${container}:${metric}:" "$WATCHDOG_STATE_FILE" 2>/dev/null | cut -d':' -f3
|
|
}
|
|
|
|
# Sets strike count for a container/metric pair.
|
|
# Usage: set_strikes "Emby" "CPU" 2
|
|
set_strikes() {
|
|
local container="$1" metric="$2" count="$3"
|
|
grep -vE "^${container}:${metric}:" "$WATCHDOG_STATE_FILE" 2>/dev/null > "${WATCHDOG_STATE_FILE}.tmp"
|
|
echo "${container}:${metric}:${count}" >> "${WATCHDOG_STATE_FILE}.tmp"
|
|
mv "${WATCHDOG_STATE_FILE}.tmp" "$WATCHDOG_STATE_FILE"
|
|
}
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# HELPERS
|
|
# -----------------------------------------------------------------------------------------------
|
|
|
|
# Fetches memory and CPU stats for a container in a single docker stats call.
|
|
# Returns: MEM_USAGE|CPU_PERCENT
|
|
parse_stats() {
|
|
local container="$1"
|
|
docker stats --no-stream --format "{{.MemUsage}}|{{.CPUPerc}}" "$container"
|
|
}
|
|
|
|
# Converts a memory value and unit to MB.
|
|
# Supports KiB, MiB, GiB — returns UNKNOWN for unrecognised units.
|
|
convert_to_mb() {
|
|
local value="$1" unit="$2"
|
|
case "$unit" in
|
|
KiB) awk "BEGIN {print $value / 1024}" ;;
|
|
MiB) echo "$value" ;;
|
|
GiB) awk "BEGIN {print $value * 1024}" ;;
|
|
*) echo "UNKNOWN" ;;
|
|
esac
|
|
}
|
|
|
|
# Restarts a container locally.
|
|
# In dry run mode reports what would happen without acting.
|
|
restart_container() {
|
|
local container="$1" reason="$2"
|
|
|
|
if [[ "$DRY_RUN" == true ]]; then
|
|
warn "DRY RUN — would restart $container ($reason)"
|
|
return
|
|
fi
|
|
|
|
info "Restarting $container ($reason)..."
|
|
if docker restart "$container" >/dev/null 2>&1; then
|
|
echo "$ICON_STARTED $container restarted"
|
|
log "Restarted $container — reason: $reason"
|
|
else
|
|
error "Failed to restart $container"
|
|
fi
|
|
}
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# MEMORY CHECK
|
|
# Compares current container memory usage against its configured hard limit.
|
|
# Restarts immediately if at or above 100% of limit.
|
|
# Warns if at or above SOFT_MEM_THRESHOLD % of limit.
|
|
# Usage: check_memory "Emby" 16384
|
|
# -----------------------------------------------------------------------------------------------
|
|
check_memory() {
|
|
local container="$1" limit_mb="$2"
|
|
local stats mem_raw mem_val mem_unit mem_mb usage_pct
|
|
|
|
stats=$(parse_stats "$container")
|
|
mem_raw=$(echo "$stats" | awk -F'|' '{print $1}' | awk '{print $1}')
|
|
mem_val=$(echo "$mem_raw" | sed -E 's/([0-9.]+).*/\1/')
|
|
mem_unit=$(echo "$mem_raw" | sed -E 's/[0-9.]+([a-zA-Z]+).*/\1/')
|
|
mem_mb=$(convert_to_mb "$mem_val" "$mem_unit")
|
|
local mem_int
|
|
mem_int=$(printf "%.0f" "$mem_mb")
|
|
usage_pct=$(( (mem_int * 100) / limit_mb ))
|
|
|
|
if (( usage_pct >= 100 )); then
|
|
error "$ICON_MEM $container memory ${mem_int}MB (${usage_pct}%) — exceeded ${limit_mb}MB hard limit"
|
|
restart_container "$container" "memory hard limit"
|
|
set_strikes "$container" "CPU" 0
|
|
set_strikes "$container" "RESP" 0
|
|
elif (( usage_pct >= SOFT_MEM_THRESHOLD )); then
|
|
warn "$ICON_MEM $container memory ${mem_int}MB (${usage_pct}% of ${limit_mb}MB limit)"
|
|
else
|
|
success "$ICON_MEM $container memory ${mem_int}MB (${usage_pct}% of ${limit_mb}MB limit)"
|
|
fi
|
|
}
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# CPU CHECK
|
|
# Normalises CPU usage against total core count and applies the strike system.
|
|
# Warns at SOFT_CPU_THRESHOLD, strikes at HARD_CPU_THRESHOLD.
|
|
# Restarts after CPU_FAIL_LIMIT consecutive strikes — resets strikes on restart or recovery.
|
|
# Usage: check_cpu "Emby"
|
|
# -----------------------------------------------------------------------------------------------
|
|
check_cpu() {
|
|
local container="$1"
|
|
local stats cpu_raw cpu_norm cpu_int violations
|
|
|
|
stats=$(parse_stats "$container")
|
|
cpu_raw=$(echo "$stats" | awk -F'|' '{print $2}' | tr -d '%')
|
|
cpu_norm=$(awk "BEGIN {print $cpu_raw / $TOTAL_CORES}")
|
|
cpu_int=$(printf "%.0f" "$cpu_norm")
|
|
|
|
violations=$(get_strikes "$container" "CPU")
|
|
[[ -z "$violations" ]] && violations=0
|
|
|
|
if (( cpu_int >= HARD_CPU_THRESHOLD )); then
|
|
((violations++))
|
|
error "$ICON_ZFS $container CPU ${cpu_int}% — hard threshold hit ($violations/$CPU_FAIL_LIMIT strikes)"
|
|
set_strikes "$container" "CPU" "$violations"
|
|
elif (( cpu_int >= SOFT_CPU_THRESHOLD )); then
|
|
((violations++))
|
|
warn "$ICON_ZFS $container CPU ${cpu_int}% — soft threshold hit ($violations/$CPU_FAIL_LIMIT strikes)"
|
|
set_strikes "$container" "CPU" "$violations"
|
|
else
|
|
if (( violations > 0 )); then
|
|
info "$ICON_ZFS $container CPU ${cpu_int}% — recovered, resetting strikes"
|
|
else
|
|
success "$ICON_ZFS $container CPU ${cpu_int}%"
|
|
fi
|
|
set_strikes "$container" "CPU" 0
|
|
violations=0
|
|
fi
|
|
|
|
if (( violations >= CPU_FAIL_LIMIT )); then
|
|
error "$ICON_ZFS $container hit CPU limit for $CPU_FAIL_LIMIT consecutive checks"
|
|
restart_container "$container" "sustained CPU abuse"
|
|
set_strikes "$container" "CPU" 0
|
|
fi
|
|
}
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# RESPONSIVENESS CHECK
|
|
# Sends an HTTP request to the container's configured URL.
|
|
# Skips containers with no URL defined in WATCHDOG_CONTAINER_URLS.
|
|
# Applies the same strike system as CPU — restarts after RESP_FAIL_LIMIT consecutive failures.
|
|
# Usage: check_responsiveness "Emby"
|
|
# -----------------------------------------------------------------------------------------------
|
|
check_responsiveness() {
|
|
local container="$1"
|
|
local url="${WATCHDOG_CONTAINER_URLS[$container]:-}"
|
|
|
|
[[ -z "$url" ]] && return
|
|
|
|
local fails
|
|
fails=$(get_strikes "$container" "RESP")
|
|
[[ -z "$fails" ]] && fails=0
|
|
|
|
if ! curl -s --max-time "$CURL_TIMEOUT" "$url" >/dev/null 2>&1; then
|
|
((fails++))
|
|
warn "$ICON_PING $container unresponsive at $url ($fails/$RESP_FAIL_LIMIT strikes)"
|
|
set_strikes "$container" "RESP" "$fails"
|
|
else
|
|
if (( fails > 0 )); then
|
|
info "$ICON_PING $container responsive again — resetting strikes"
|
|
else
|
|
success "$ICON_PING $container responsive at $url"
|
|
fi
|
|
set_strikes "$container" "RESP" 0
|
|
fails=0
|
|
fi
|
|
|
|
if (( fails >= RESP_FAIL_LIMIT )); then
|
|
error "$ICON_PING $container unresponsive for $RESP_FAIL_LIMIT consecutive checks"
|
|
restart_container "$container" "HTTP unresponsive"
|
|
set_strikes "$container" "RESP" 0
|
|
fi
|
|
}
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# ━━━ $ICON_WATCHDOG Watchdog Check ━━━
|
|
# -----------------------------------------------------------------------------------------------
|
|
echo ""
|
|
echo "━━━ $ICON_WATCHDOG Watchdog Check — $(date '+%Y-%m-%d %H:%M:%S') ━━━"
|
|
echo ""
|
|
|
|
RESTARTED=()
|
|
SKIPPED=()
|
|
|
|
START=$(date +%s)
|
|
|
|
for container in "${!WATCHDOG_CONTAINERS[@]}"; do
|
|
echo "━━━ $ICON_CONTAINERS $container ━━━"
|
|
|
|
# Verify container exists
|
|
if ! docker inspect "$container" &>/dev/null; then
|
|
warn "$container not found on this host — skipping"
|
|
SKIPPED+=("$container")
|
|
echo ""
|
|
continue
|
|
fi
|
|
|
|
# Verify container is running
|
|
if ! docker ps --filter "name=^/${container}$" --format "{{.Names}}" | grep -qw "$container"; then
|
|
warn "$ICON_NOT_RUNNING $container is not running — skipping"
|
|
SKIPPED+=("$container")
|
|
echo ""
|
|
continue
|
|
fi
|
|
|
|
check_memory "$container" "${WATCHDOG_CONTAINERS[$container]}"
|
|
check_cpu "$container"
|
|
check_responsiveness "$container"
|
|
|
|
echo ""
|
|
done
|
|
|
|
END=$(date +%s)
|
|
|
|
# -----------------------------------------------------------------------------------------------
|
|
# ━━━ $ICON_SUMMARY Summary ━━━
|
|
# -----------------------------------------------------------------------------------------------
|
|
echo "━━━━━ $ICON_SUMMARY WATCHDOG SUMMARY ━━━━━"
|
|
echo "$ICON_TIME $(date '+%Y-%m-%d %H:%M:%S')"
|
|
echo "$ICON_TIME Duration: $(format_duration $((END - START)))"
|
|
echo "$ICON_WATCHDOG Monitored: ${#WATCHDOG_CONTAINERS[@]} containers"
|
|
echo "$ICON_NOT_RUNNING Skipped: ${#SKIPPED[@]} containers"
|
|
if [[ "$DRY_RUN" == true ]]; then
|
|
echo "$ICON_WARN Dry Run: no restarts executed"
|
|
fi
|
|
echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━" |