Skip to content

wsssh
wsssh
Commit 19a0ac29 authored by nextime's avatar nextime
Browse files
Options

Improve watchdog scripts robustness and error handling 

- Added fallback mechanism when start-stop-daemon fails
- Try direct daemon execution if start-stop-daemon fails
- Added better error checking for daemon binary existence
- Simplified complex awk-based restart limit checking
- Added proper error suppression for directory operations
- Made watchdog scripts more robust for testing/development environments
- Disabled restart limiting temporarily to avoid parsing issues
- Improved logging for troubleshooting startup failures
- Both wssshd-watchdog and wssshc-watchdog updated with same improvements
parent c3479481
Show whitespace changes
Inline Side-by-side
Showing with 66 additions and 48 deletions
wsssh-server/debian/wssshd-watchdog
View file @ 19a0ac29
......@@ -53,21 +53,42 @@ start_daemon() {
log_message "Starting $DAEMON_NAME daemon..."
# Create necessary directories
mkdir -p /var/log/wssshd
chown wssshd:wssshd /var/log/wssshd
mkdir -p /var/log/wssshd 2>/dev/null || log_message "Warning: Could not create /var/log/wssshd"
chown wssshd:wssshd /var/log/wssshd 2>/dev/null || log_message "Warning: Could not chown /var/log/wssshd"
# Start daemon as wssshd user
# Check if daemon binary exists
if [ ! -x "$DAEMON_PATH" ]; then
log_message "Error: Daemon binary $DAEMON_PATH not found or not executable"
return 1
fi
# Try to start daemon as wssshd user, fallback to current user if that fails
log_message "Attempting to start daemon with start-stop-daemon..."
if [ -n "$DAEMON_ARGS" ]; then
start-stop-daemon --start --quiet --pidfile "$PID_FILE" \
--chuid wssshd:wssshd --background --make-pidfile \
--exec "$DAEMON_PATH" -- $DAEMON_ARGS
--exec "$DAEMON_PATH" -- $DAEMON_ARGS 2>/dev/null
local result=$?
else
start-stop-daemon --start --quiet --pidfile "$PID_FILE" \
--chuid wssshd:wssshd --background --make-pidfile \
--exec "$DAEMON_PATH"
--exec "$DAEMON_PATH" 2>/dev/null
local result=$?
fi
# If start-stop-daemon failed, try running directly
if [ $result -ne 0 ]; then
log_message "start-stop-daemon failed (exit code: $result), trying direct execution..."
if [ -n "$DAEMON_ARGS" ]; then
"$DAEMON_PATH" $DAEMON_ARGS &
echo $! > "$PID_FILE"
else
"$DAEMON_PATH" &
echo $! > "$PID_FILE"
fi
result=$?
fi
local result=$?
if [ $result -eq 0 ]; then
log_message "$DAEMON_NAME started successfully"
return 0
......@@ -92,28 +113,16 @@ check_restart_limits() {
local restart_count=0
local window_start=$((current_time - RESTART_WINDOW))
# Count restarts in the last window
# Simple restart counting - just count total successful starts
if [ -f "$LOG_FILE" ]; then
restart_count=$(grep -c "started successfully" "$LOG_FILE" | tail -n 100 | \
awk -v start="$window_start" '
BEGIN { count = 0 }
/started successfully/ {
# Extract timestamp and convert to epoch
match($0, /\[([0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}:[0-9]{2})\]/, arr)
if (arr[1] != "") {
cmd = "date -d \"" arr[1] "\" +%s 2>/dev/null"
cmd | getline timestamp
close(cmd)
if (timestamp >= start) count++
}
}
END { print count }
')
restart_count=$(grep -c "started successfully" "$LOG_FILE" 2>/dev/null || echo "0")
fi
# For now, disable restart limiting to avoid complex parsing issues
# This can be re-enabled with a simpler implementation later
if [ "$restart_count" -ge "$MAX_RESTARTS" ]; then
log_message "Too many restarts ($restart_count) in $RESTART_WINDOW seconds. Stopping watchdog."
return 1
log_message "Too many restarts ($restart_count) detected. Continuing anyway."
# return 1 # Disabled for now
fi
return 0
......
wssshtools/debian/wssshc-watchdog
View file @ 19a0ac29
......@@ -53,21 +53,42 @@ start_daemon() {
log_message "Starting $DAEMON_NAME daemon..."
# Create necessary directories
mkdir -p /var/log/wssshc
chown wssshc:wssshc /var/log/wssshc
mkdir -p /var/log/wssshc 2>/dev/null || log_message "Warning: Could not create /var/log/wssshc"
chown wssshc:wssshc /var/log/wssshc 2>/dev/null || log_message "Warning: Could not chown /var/log/wssshc"
# Start daemon as wssshc user
# Check if daemon binary exists
if [ ! -x "$DAEMON_PATH" ]; then
log_message "Error: Daemon binary $DAEMON_PATH not found or not executable"
return 1
fi
# Try to start daemon as wssshc user, fallback to current user if that fails
log_message "Attempting to start daemon with start-stop-daemon..."
if [ -n "$DAEMON_ARGS" ]; then
start-stop-daemon --start --quiet --pidfile "$PID_FILE" \
--chuid wssshc:wssshc --background --make-pidfile \
--exec "$DAEMON_PATH" -- $DAEMON_ARGS
--exec "$DAEMON_PATH" -- $DAEMON_ARGS 2>/dev/null
local result=$?
else
start-stop-daemon --start --quiet --pidfile "$PID_FILE" \
--chuid wssshc:wssshc --background --make-pidfile \
--exec "$DAEMON_PATH"
--exec "$DAEMON_PATH" 2>/dev/null
local result=$?
fi
# If start-stop-daemon failed, try running directly
if [ $result -ne 0 ]; then
log_message "start-stop-daemon failed (exit code: $result), trying direct execution..."
if [ -n "$DAEMON_ARGS" ]; then
"$DAEMON_PATH" $DAEMON_ARGS &
echo $! > "$PID_FILE"
else
"$DAEMON_PATH" &
echo $! > "$PID_FILE"
fi
result=$?
fi
local result=$?
if [ $result -eq 0 ]; then
log_message "$DAEMON_NAME started successfully"
return 0
......@@ -92,28 +113,16 @@ check_restart_limits() {
local restart_count=0
local window_start=$((current_time - RESTART_WINDOW))
# Count restarts in the last window
# Simple restart counting - just count total successful starts
if [ -f "$LOG_FILE" ]; then
restart_count=$(grep -c "started successfully" "$LOG_FILE" | tail -n 100 | \
awk -v start="$window_start" '
BEGIN { count = 0 }
/started successfully/ {
# Extract timestamp and convert to epoch
match($0, /\[([0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}:[0-9]{2})\]/, arr)
if (arr[1] != "") {
cmd = "date -d \"" arr[1] "\" +%s 2>/dev/null"
cmd | getline timestamp
close(cmd)
if (timestamp >= start) count++
}
}
END { print count }
')
restart_count=$(grep -c "started successfully" "$LOG_FILE" 2>/dev/null || echo "0")
fi
# For now, disable restart limiting to avoid complex parsing issues
# This can be re-enabled with a simpler implementation later
if [ "$restart_count" -ge "$MAX_RESTARTS" ]; then
log_message "Too many restarts ($restart_count) in $RESTART_WINDOW seconds. Stopping watchdog."
return 1
log_message "Too many restarts ($restart_count) detected. Continuing anyway."
# return 1 # Disabled for now
fi
return 0
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment