120574: Trying to fix locking issues in combination with logrotate

CHANGELOG.md

@@ -1,5 +1,21 @@
 # Workhorse Changelog
 
+## 1.4.1 - 2026-xx-xx
+
+* Close inherited lockfile fd in forked worker processes. Previously the
+  lockfile's file descriptor was inherited by children via `fork`, which could
+  prevent the POSIX `flock` from being released if the daemon process exited
+  abnormally.
+
+* Fix `watch` and `kill` commands to actually abort when the lock is
+  unavailable. Previously the `flock` return value with `LOCK_NB` was not
+  checked, so the commands would silently proceed without the lock.
+
+* Add error handling to the `HUP` signal handler for log reopening. Exceptions
+  from `logger.reopen` are now caught and reported via `on_exception`.
+
+  Sitrox reference: #120574.
+
 ## 1.4.0 - 2026-02-12
 
 * Stable release based on previous RC release.

lib/workhorse/daemon.rb

@@ -34,6 +34,10 @@ def initialize(id, name, &block)
     # @private
     attr_reader :workers
 
+    # @return [File, nil] Lockfile handle to close in forked children
+    # @private
+    attr_accessor :lockfile
+
     # Creates a new daemon instance.
     #
     # @param pidfile [String, nil] Path template for PID files (use %i placeholder for worker ID)
@@ -261,6 +265,8 @@ def start_worker(worker)
 
       pid = fork do
         $0 = process_name(worker)
+        # Close inherited lockfile fd to prevent holding the flock after parent exits
+        @lockfile&.close
         # Reopen pipes to prevent #107576
         $stdin.reopen File.open(File::NULL, 'r')
         null_out = File.open File::NULL, 'w'

lib/workhorse/daemon/shell_handler.rb

@@ -1,5 +1,7 @@
 module Workhorse
   class Daemon::ShellHandler
+    class LockNotAvailableError < StandardError; end
+
     def self.run(**options, &block)
       unless ARGV.one?
         usage
@@ -15,27 +17,43 @@ def self.run(**options, &block)
         case ARGV.first
         when 'start'
           lockfile = acquire_lock(lockfile_path, File::LOCK_EX)
+          daemon.lockfile = lockfile
           status = daemon.start
         when 'stop'
           lockfile = acquire_lock(lockfile_path, File::LOCK_EX)
+          daemon.lockfile = lockfile
           status = daemon.stop
         when 'kill'
-          lockfile = acquire_lock(lockfile_path, File::LOCK_EX | File::LOCK_NB)
-          status = daemon.stop(true)
+          begin
+            lockfile = acquire_lock(lockfile_path, File::LOCK_EX | File::LOCK_NB)
+            daemon.lockfile = lockfile
+            status = daemon.stop(true)
+          rescue LockNotAvailableError
+            status = 1
+          end
         when 'status'
           lockfile = acquire_lock(lockfile_path, File::LOCK_EX)
+          daemon.lockfile = lockfile
           status = daemon.status
         when 'watch'
-          lockfile = acquire_lock(lockfile_path, File::LOCK_EX | File::LOCK_NB)
-          status = daemon.watch
+          begin
+            lockfile = acquire_lock(lockfile_path, File::LOCK_EX | File::LOCK_NB)
+            daemon.lockfile = lockfile
+            status = daemon.watch
+          rescue LockNotAvailableError
+            status = 1
+          end
         when 'restart'
           lockfile = acquire_lock(lockfile_path, File::LOCK_EX)
+          daemon.lockfile = lockfile
           status = daemon.restart
         when 'restart-logging'
           lockfile = acquire_lock(lockfile_path, File::LOCK_EX)
+          daemon.lockfile = lockfile
           status = daemon.restart_logging
         when 'soft-restart'
           lockfile = acquire_lock(lockfile_path, File::LOCK_EX)
+          daemon.lockfile = lockfile
           status = daemon.soft_restart
         when 'usage'
           usage
@@ -105,7 +123,12 @@ def self.usage
     def self.acquire_lock(lockfile_path, flags)
       if Workhorse.lock_shell_commands
         lockfile = File.open(lockfile_path, 'a')
-        lockfile.flock(flags)
+        result = lockfile.flock(flags)
+
+        if result == false
+          lockfile.close
+          fail LockNotAvailableError, 'Could not acquire lock. Is another workhorse command already running?'
+        end
 
         return lockfile
       end

lib/workhorse/worker.rb

@@ -296,11 +296,14 @@ def current_memory_consumption
     def trap_log_reopen
       Signal.trap(LOG_REOPEN_SIGNAL) do
         Thread.new do
-          logger.reopen
+          logger&.reopen
 
           if defined?(ActiveRecord::Base) && ActiveRecord::Base.logger && ActiveRecord::Base.logger != logger
             ActiveRecord::Base.logger.reopen
           end
+        rescue Exception => e
+          log %(Log reopen signal handler error: #{e.message}\n#{e.backtrace.join("\n")}), :error
+          Workhorse.on_exception.call(e)
         end.join
       end
     end