InnoDB: Increment the lock wait watchdog timeout during CHECK TABLE

(Bug #2694)


innobase/include/srv0srv.h:
  Add srv_fatal_semaphore_wait_threshold
innobase/include/sync0arr.h:
  Improve comment of sync_array_print_long_waits()
innobase/row/row0mysql.c:
  Lengthen the srv_fatal_semaphore_wait_threshold by 2 hours during
  CHECK TABLE
innobase/srv/srv0srv.c:
  Add srv_fatal_semaphore_wait_threshold
innobase/sync/sync0arr.c:
  Improve comment of sync_array_print_long_waits().
  Replace the fixed timeout of 600 seconds with
  srv_fatal_semaphore_wait_threshold.

innobase/include/srv0srv.h

@@ -149,6 +149,7 @@ extern	ulint	srv_test_n_mutexes;
 extern	ulint	srv_test_array_size;
 
 extern ulint	srv_activity_count;
+extern ulint	srv_fatal_semaphore_wait_threshold;
 
 extern mutex_t*	kernel_mutex_temp;/* mutex protecting the server, trx structs,
 				query threads, and lock table: we allocate

innobase/include/sync0arr.h

@@ -95,7 +95,7 @@ void
 sync_arr_wake_threads_if_sema_free(void);
 /*====================================*/
 /**************************************************************************
-Prints warnings of long semaphore waits to stderr. Currently > 120 sec. */
+Prints warnings of long semaphore waits to stderr. */
 
 void
 sync_array_print_long_waits(void);

innobase/row/row0mysql.c

@@ -2805,7 +2805,12 @@ row_check_table_for_mysql(
 	REPEATABLE READ here */
 
 	prebuilt->trx->isolation_level = TRX_ISO_REPEATABLE_READ;
-	
+
+	/* Enlarge the fatal lock wait timeout during CHECK TABLE. */
+	mutex_enter(&kernel_mutex);
+	srv_fatal_semaphore_wait_threshold += 7200; /* 2 hours */
+	mutex_exit(&kernel_mutex);
+
 	index = dict_table_get_first_index(table);
 
 	while (index != NULL) {
@@ -2853,6 +2858,11 @@ row_check_table_for_mysql(
 		ret = DB_ERROR;
 	}
 
+	/* Restore the fatal lock wait timeout after CHECK TABLE. */
+	mutex_enter(&kernel_mutex);
+	srv_fatal_semaphore_wait_threshold -= 7200; /* 2 hours */
+	mutex_exit(&kernel_mutex);
+
 	prebuilt->trx->op_info = (char *) "";
 
 	return(ret);

innobase/srv/srv0srv.c

@@ -55,6 +55,9 @@ ibool	srv_lower_case_table_names	= FALSE;
 in the server */
 ulint	srv_activity_count	= 0;
 
+/* The following is the maximum allowed duration of a lock wait. */
+ulint	srv_fatal_semaphore_wait_threshold = 600;
+
 ibool	srv_lock_timeout_and_monitor_active = FALSE;
 ibool	srv_error_monitor_active = FALSE;
 

innobase/sync/sync0arr.c

@@ -890,7 +890,7 @@ sync_arr_wake_threads_if_sema_free(void)
 }
 
 /**************************************************************************
-Prints warnings of long semaphore waits to stderr. Currently > 120 sec. */
+Prints warnings of long semaphore waits to stderr. */
 
 void
 sync_array_print_long_waits(void)
@@ -900,6 +900,7 @@ sync_array_print_long_waits(void)
         ibool		old_val;
 	ibool		noticed = FALSE;
 	ulint           i;
+	ulint		fatal_timeout = srv_fatal_semaphore_wait_threshold;
 
         for (i = 0; i < sync_primary_wait_array->n_cells; i++) {
 
@@ -914,12 +915,13 @@ sync_array_print_long_waits(void)
                 }
 
                 if (cell->wait_object != NULL
-		    && difftime(time(NULL), cell->reservation_time) > 600) {
+		    && difftime(time(NULL), cell->reservation_time)
+		    > fatal_timeout) {
 
-			fputs(
-"InnoDB: Error: semaphore wait has lasted > 600 seconds\n"
+			fprintf(stderr,
+"InnoDB: Error: semaphore wait has lasted > %lu seconds\n"
 "InnoDB: We intentionally crash the server, because it appears to be hung.\n",
-				stderr);
+				fatal_timeout);
 
 		    	ut_error;
                 }