MDEV-20220: Merge 5.7 P_S replication table 'replication_applier_status_by_worker

Step2:
=====
Add two extra columns mentioned below.
---------------------------------------------------------------------------
|Column Name:           |        Description:                             |
|-------------------------------------------------------------------------|
|                       |                                                 |
|WORKER_IDLE_TIME       | Total idle time in seconds that the worker      |
|                       | thread has spent waiting for work from          |
|                       | co-ordinator thread                             |
|                       |                                                 |
|LAST_TRANS_RETRY_COUNT | Total number of retries attempted by last       |
|                       | transaction                                     |
---------------------------------------------------------------------------

mysql-test/suite/perfschema/r/table_schema.result

@@ -872,6 +872,8 @@ def	performance_schema	replication_applier_status_by_worker	LAST_SEEN_TRANSACTIO
 def	performance_schema	replication_applier_status_by_worker	LAST_ERROR_NUMBER	5	NULL	NO	int	NULL	NULL	10	0	NULL	NULL	NULL	int(11)			select,insert,update,references		NEVER	NULL
 def	performance_schema	replication_applier_status_by_worker	LAST_ERROR_MESSAGE	6	NULL	NO	varchar	1024	3072	NULL	NULL	NULL	utf8	utf8_general_ci	varchar(1024)			select,insert,update,references		NEVER	NULL
 def	performance_schema	replication_applier_status_by_worker	LAST_ERROR_TIMESTAMP	7	current_timestamp()	NO	timestamp	NULL	NULL	NULL	NULL	0	NULL	NULL	timestamp		on update current_timestamp()	select,insert,update,references		NEVER	NULL
+def	performance_schema	replication_applier_status_by_worker	WORKER_IDLE_TIME	8	NULL	NO	bigint	NULL	NULL	20	0	NULL	NULL	NULL	bigint(20) unsigned			select,insert,update,references		NEVER	NULL
+def	performance_schema	replication_applier_status_by_worker	LAST_TRANS_RETRY_COUNT	9	NULL	NO	int	NULL	NULL	10	0	NULL	NULL	NULL	int(11)			select,insert,update,references		NEVER	NULL
 def	performance_schema	replication_connection_configuration	CHANNEL_NAME	1	NULL	NO	char	64	192	NULL	NULL	NULL	utf8	utf8_general_ci	char(64)			select,insert,update,references		NEVER	NULL
 def	performance_schema	replication_connection_configuration	HOST	2	NULL	NO	char	60	180	NULL	NULL	NULL	utf8	utf8_bin	char(60)			select,insert,update,references		NEVER	NULL
 def	performance_schema	replication_connection_configuration	PORT	3	NULL	NO	int	NULL	NULL	10	0	NULL	NULL	NULL	int(11)			select,insert,update,references		NEVER	NULL

mysql-test/suite/rpl/r/rpl_parallel_retry.result

@@ -49,8 +49,7 @@ SET @old_dbug= @@GLOBAL.debug_dbug;
 SET GLOBAL debug_dbug="+d,rpl_parallel_simulate_temp_err_gtid_0_x_100";
 include/start_slave.inc
 SET GLOBAL debug_dbug=@old_dbug;
-retries
-1
+include/assert.inc [Performance Schema retries should match with actual retries]
 SELECT * FROM t1 ORDER BY a;
 a	b
 1	2
@@ -82,8 +81,7 @@ SET @old_dbug= @@GLOBAL.debug_dbug;
 SET GLOBAL debug_dbug="+d,rpl_parallel_simulate_temp_err_gtid_0_x_100,rpl_parallel_simulate_double_temp_err_gtid_0_x_100";
 include/start_slave.inc
 SET GLOBAL debug_dbug=@old_dbug;
-retries
-2
+include/assert.inc [Performance Schema retries should match with actual retries]
 SELECT * FROM t1 ORDER BY a;
 a	b
 1	3
@@ -179,8 +177,7 @@ SET @old_dbug= @@GLOBAL.debug_dbug;
 SET GLOBAL debug_dbug="+d,rpl_parallel_simulate_temp_err_gtid_0_x_100";
 include/start_slave.inc
 SET GLOBAL debug_dbug=@old_dbug;
-retries
-1
+include/assert.inc [Performance Schema retries should match with actual retries]
 SELECT * FROM t1 WHERE a >= 10 ORDER BY a;
 a	b
 10	4
@@ -224,8 +221,7 @@ SET @old_dbug= @@GLOBAL.debug_dbug;
 SET GLOBAL debug_dbug="+d,rpl_parallel_simulate_temp_err_xid";
 include/start_slave.inc
 SET GLOBAL debug_dbug=@old_dbug;
-retries
-1
+include/assert.inc [Performance Schema retries should match with actual retries]
 SELECT * FROM t1 WHERE a >= 100 ORDER BY a;
 a	b
 100	0
@@ -323,8 +319,7 @@ SET debug_sync='now WAIT_FOR t3_waiting';
 SET debug_sync='now SIGNAL t1_start';
 SET GLOBAL debug_dbug=@old_dbug;
 SET debug_sync='RESET';
-retries
-1
+include/assert.inc [Performance Schema retries should match with actual retries]
 SELECT * FROM t3 ORDER BY a;
 a	b
 1	NULL

mysql-test/suite/rpl/r/rpl_perfschema_applier_status_by_worker.result

@@ -28,6 +28,9 @@ include/assert.inc [Value returned by SSS and PS table for Last_Error_Number sho
 include/assert.inc [Value returned by SSS and PS table for Last_Error_Message should both be empty.]
 include/assert.inc [Value returned by PS table for Last_Error_Timestamp should be 0000-00-00 00:00:00.]
 connection master;
+connection slave;
+include/assert.inc [Value returned by PS table for worker_idle_time should be >= 1]
+connection master;
 DROP TABLE t1;
 connection slave;
 STOP SLAVE 'slave1';

mysql-test/suite/rpl/t/rpl_parallel_retry.test

@@ -68,9 +68,13 @@ let $old_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', V
 --sync_with_master
 SET GLOBAL debug_dbug=@old_dbug;
 let $new_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1);
---disable_query_log
-eval SELECT $new_retry - $old_retry AS retries;
---enable_query_log
+let $ps_value= query_get_value(select last_trans_retry_count from
+    performance_schema.replication_applier_status_by_worker where
+    last_trans_retry_count > 0, last_trans_retry_count, 1);
+let $assert_text= Performance Schema retries should match with actual retries;
+let $assert_cond= "$ps_value" = $new_retry - $old_retry;
+source include/assert.inc;
+
 
 SELECT * FROM t1 ORDER BY a;
 
@@ -100,9 +104,12 @@ let $old_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', V
 --sync_with_master
 SET GLOBAL debug_dbug=@old_dbug;
 let $new_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1);
---disable_query_log
-eval SELECT $new_retry - $old_retry AS retries;
---enable_query_log
+let $ps_value= query_get_value(select last_trans_retry_count from
+    performance_schema.replication_applier_status_by_worker where
+    last_trans_retry_count > 0, last_trans_retry_count, 1);
+let $assert_text= Performance Schema retries should match with actual retries;
+let $assert_cond= "$ps_value" = $new_retry - $old_retry;
+source include/assert.inc;
 
 SELECT * FROM t1 ORDER BY a;
 
@@ -187,9 +194,12 @@ let $old_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', V
 --sync_with_master
 SET GLOBAL debug_dbug=@old_dbug;
 let $new_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1);
---disable_query_log
-eval SELECT $new_retry - $old_retry AS retries;
---enable_query_log
+let $ps_value= query_get_value(select last_trans_retry_count from
+    performance_schema.replication_applier_status_by_worker where
+    last_trans_retry_count > 0, last_trans_retry_count, 1);
+let $assert_text= Performance Schema retries should match with actual retries;
+let $assert_cond= "$ps_value" = $new_retry - $old_retry;
+source include/assert.inc;
 
 SELECT * FROM t1 WHERE a >= 10 ORDER BY a;
 SELECT a, LENGTH(b) FROM t2 ORDER BY a;
@@ -235,9 +245,12 @@ let $old_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', V
 --sync_with_master
 SET GLOBAL debug_dbug=@old_dbug;
 let $new_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1);
---disable_query_log
-eval SELECT $new_retry - $old_retry AS retries;
---enable_query_log
+let $ps_value= query_get_value(select last_trans_retry_count from
+    performance_schema.replication_applier_status_by_worker where
+    last_trans_retry_count > 0, last_trans_retry_count, 1);
+let $assert_text= Performance Schema retries should match with actual retries;
+let $assert_cond= "$ps_value" = $new_retry - $old_retry;
+source include/assert.inc;
 
 SELECT * FROM t1 WHERE a >= 100 ORDER BY a;
 # Stop the SQL thread. When the bug was there to give the incorrect relay log
@@ -365,9 +378,12 @@ SET debug_sync='now SIGNAL t1_start';
 SET GLOBAL debug_dbug=@old_dbug;
 SET debug_sync='RESET';
 let $new_retry= query_get_value(SHOW STATUS LIKE 'Slave_retried_transactions', Value, 1);
---disable_query_log
-eval SELECT $new_retry - $old_retry >= 1 AS retries;
---enable_query_log
+let $ps_value= query_get_value(select last_trans_retry_count from
+    performance_schema.replication_applier_status_by_worker where
+    last_trans_retry_count > 0, last_trans_retry_count, 1);
+let $assert_text= Performance Schema retries should match with actual retries;
+let $assert_cond= "$ps_value" = $new_retry - $old_retry;
+source include/assert.inc;
 SELECT * FROM t3 ORDER BY a;
 
 

mysql-test/suite/rpl/t/rpl_perfschema_applier_status_by_worker.test

@@ -113,6 +113,14 @@ let $assert_text= Value returned by PS table for Last_Error_Timestamp should be
 let $assert_cond= "$ps_value" = "0000-00-00 00:00:00";
 source include/assert.inc;
 
+--connection master
+sleep 1;
+--connection slave
+let $ps_value= query_get_value(select worker_idle_time from performance_schema.replication_applier_status_by_worker, worker_idle_time, 1);
+let $assert_text= Value returned by PS table for worker_idle_time should be >= 1;
+let $assert_cond= "$ps_value" >= "1";
+source include/assert.inc;
+
 --connection master
 DROP TABLE t1;
 --save_master_pos

sql/rpl_parallel.cc

@@ -799,6 +799,7 @@ retry_event_group(rpl_group_info *rgi, rpl_parallel_thread *rpt,
 
   mysql_mutex_lock(&rli->data_lock);
   ++rli->retried_trans;
+  ++rpt->last_trans_retry_count;
   statistic_increment(slave_retried_transactions, LOCK_status);
   mysql_mutex_unlock(&rli->data_lock);
 
@@ -1125,6 +1126,7 @@ handle_rpl_parallel_thread(void *arg)
     uint wait_count= 0;
     rpl_parallel_thread::queued_event *qev, *next_qev;
 
+    rpt->start_time_tracker();
     thd->ENTER_COND(&rpt->COND_rpl_thread, &rpt->LOCK_rpl_thread,
                     &stage_waiting_for_work_from_sql_thread, &old_stage);
     /*
@@ -1148,6 +1150,7 @@ handle_rpl_parallel_thread(void *arg)
     }
     rpt->dequeue1(events);
     thd->EXIT_COND(&old_stage);
+    rpt->add_to_worker_idle_time_and_reset();
 
   more_events:
     for (qev= events; qev; qev= next_qev)
@@ -1193,6 +1196,7 @@ handle_rpl_parallel_thread(void *arg)
       /* Handle a new event group, which will be initiated by a GTID event. */
       if ((event_type= qev->ev->get_type_code()) == GTID_EVENT)
       {
+        rpt->last_trans_retry_count= 0;
         rpt->last_seen_gtid= rgi->current_gtid;
         rpt->channel_name_length= (uint)rgi->rli->mi->connection_name.length;
         if (rpt->channel_name_length)
@@ -2015,7 +2019,8 @@ rpl_parallel_thread::loc_free_gco(group_commit_orderer *gco)
 
 
 rpl_parallel_thread::rpl_parallel_thread()
-  : channel_name_length(0), last_error_number(0), last_error_timestamp(0)
+  : channel_name_length(0), last_error_number(0), last_error_timestamp(0),
+    worker_idle_time(0), last_trans_retry_count(0), start_time(0)
 {
 }
 

sql/rpl_parallel.h

@@ -167,6 +167,29 @@ struct rpl_parallel_thread {
   int last_error_number;
   char last_error_message[MAX_SLAVE_ERRMSG];
   ulonglong last_error_timestamp;
+  ulonglong worker_idle_time;
+  ulong last_trans_retry_count;
+  ulonglong start_time;
+  void start_time_tracker()
+  {
+    start_time= microsecond_interval_timer();
+  }
+  ulonglong compute_time_lapsed()
+  {
+    return (ulonglong)((microsecond_interval_timer() - start_time) / 1000000.0);
+  }
+  void add_to_worker_idle_time_and_reset()
+  {
+    worker_idle_time+= compute_time_lapsed();
+    start_time=0;
+  }
+  ulonglong get_worker_idle_time()
+  {
+    if (start_time)
+      return compute_time_lapsed();
+    else
+      return worker_idle_time;
+  }
   void enqueue(queued_event *qev)
   {
     if (last_in_queue)

storage/perfschema/table_replication_applier_status_by_worker.cc

@@ -58,7 +58,9 @@ table_replication_applier_status_by_worker::m_share=
   "LAST_SEEN_TRANSACTION CHAR(57) not null,"
   "LAST_ERROR_NUMBER INTEGER not null,"
   "LAST_ERROR_MESSAGE VARCHAR(1024) not null,"
-  "LAST_ERROR_TIMESTAMP TIMESTAMP(0) not null)") },
+  "LAST_ERROR_TIMESTAMP TIMESTAMP(0) not null,"
+  "WORKER_IDLE_TIME BIGINT UNSIGNED not null,"
+  "LAST_TRANS_RETRY_COUNT INTEGER not null)") },
   false  /* perpetual */
 };
 
@@ -153,6 +155,7 @@ void table_replication_applier_status_by_worker::make_row(rpl_parallel_thread *r
   if (m_row.channel_name_length)
     memcpy(m_row.channel_name, rpt->channel_name, m_row.channel_name_length);
 
+  m_row.thread_id_is_null= true;
   if (rpt->running)
   {
     PSI_thread *psi= thd_get_psi(rpt->thd);
@@ -162,11 +165,7 @@ void table_replication_applier_status_by_worker::make_row(rpl_parallel_thread *r
       m_row.thread_id= pfs->m_thread_internal_id;
       m_row.thread_id_is_null= false;
     }
-    else
-      m_row.thread_id_is_null= true;
   }
-  else
-    m_row.thread_id_is_null= true;
 
   if ((gtid.seq_no > 0 &&
        !rpl_slave_state_tostring_helper(&str, &gtid, &first)))
@@ -197,6 +196,8 @@ void table_replication_applier_status_by_worker::make_row(rpl_parallel_thread *r
     m_row.last_error_timestamp= rpt->last_error_timestamp;
   }
 
+  m_row.last_trans_retry_count= rpt->last_trans_retry_count;
+  m_row.worker_idle_time= rpt->get_worker_idle_time();
   m_row_exists= true;
 }
 
@@ -242,6 +243,12 @@ int table_replication_applier_status_by_worker
       case 6: /*last_error_timestamp*/
         set_field_timestamp(f, m_row.last_error_timestamp);
         break;
+      case 7: /*worker_idle_time*/
+        set_field_ulonglong(f, m_row.worker_idle_time);
+        break;
+      case 8: /*last_trans_retry_count*/
+        set_field_ulong(f, m_row.last_trans_retry_count);
+        break;
       default:
         DBUG_ASSERT(false);
       }

storage/perfschema/table_replication_applier_status_by_worker.h

@@ -68,6 +68,8 @@ struct st_row_worker {
   char last_error_message[MAX_SLAVE_ERRMSG];
   uint last_error_message_length;
   ulonglong last_error_timestamp;
+  ulonglong worker_idle_time;
+  ulong last_trans_retry_count;
 };
 
 /**