Bug #102079 Handler_* counters are not incremented for parallel COUNT(*)

Description:
The handler_ status counters are used for monitoring, for example, to detect full table scans, but these counters are not incremented when a COUNT(*) without a WHERE clause is used on an InnoDB table in MySQL 8.  

I imagine this problem is related to the optimized COUNT(*) implementation that is used for InnoDB (parallel implementation) on MySQL 8.  When MyISAM uses fast count(*) it does not read from the table, so it shouldn't increment status counters for COUNT(*) w/out a WHERE clause.  InnoDB, however, still does read from the table, but the counters are not updated.

How to repeat:
mysql> drop database if exists test;
Query OK, 1 row affected (0.56 sec)

mysql> create database test;
Query OK, 1 row affected (0.00 sec)

mysql> use test;
Database changed

mysql> create table handler_test(c1 int primary key) engine=innodb;
Query OK, 0 rows affected (0.15 sec)

mysql> insert into handler_test values (1),(2),(3),(4),(5),(6),(7),(8),(9),(10);
Query OK, 10 rows affected (0.01 sec)
Records: 10  Duplicates: 0  Warnings: 0

mysql> flush status;
Query OK, 0 rows affected (0.00 sec)

-- should increment status variables (faster count(*) does read rows!)
mysql> select count(*) from handler_test;
+----------+
| count(*) |
+----------+
|       10 |
+----------+
1 row in set (0.00 sec)

-- status variables are not changed
mysql> show session status like '%handler%read%';
+-----------------------+-------+
| Variable_name         | Value |
+-----------------------+-------+
| Handler_read_first    | 0     |
| Handler_read_key      | 0     |
| Handler_read_last     | 0     |
| Handler_read_next     | 0     |
| Handler_read_prev     | 0     |
| Handler_read_rnd      | 0     |
| Handler_read_rnd_next | 0     |
+-----------------------+-------+
7 rows in set (0.02 sec)

-- should increment status variables
mysql> select count(*) from handler_test where c1 > 0;
+----------+
| count(*) |
+----------+
|       10 |
+----------+
1 row in set (0.00 sec)

-- status variables ARE incremented
mysql> show session status like '%handler%read%';
+-----------------------+-------+
| Variable_name         | Value |
+-----------------------+-------+
| Handler_read_first    | 1     |
| Handler_read_key      | 1     |
| Handler_read_last     | 0     |
| Handler_read_next     | 10    |
| Handler_read_prev     | 0     |
| Handler_read_rnd      | 0     |
| Handler_read_rnd_next | 0     |
+-----------------------+-------+
7 rows in set (0.02 sec)

mysql> alter table handler_test engine=myisam;
Query OK, 10 rows affected (0.05 sec)
Records: 10  Duplicates: 0  Warnings: 0

mysql> flush status;
Query OK, 0 rows affected (0.00 sec)

-- status variables SHOULD NOT incremented (fast count(*) reads no rows)
mysql> select count(*) from handler_test;
+----------+
| count(*) |
+----------+
|       10 |
+----------+
1 row in set (0.00 sec)

-- results are correct
mysql> show session status like '%handler%read%';
+-----------------------+-------+
| Variable_name         | Value |
+-----------------------+-------+
| Handler_read_first    | 0     |
| Handler_read_key      | 0     |
| Handler_read_last     | 0     |
| Handler_read_next     | 0     |
| Handler_read_prev     | 0     |
| Handler_read_rnd      | 0     |
| Handler_read_rnd_next | 0     |
+-----------------------+-------+
7 rows in set (0.02 sec)

-- should increment status variables
mysql> select count(*) from handler_test where c1 > 0;
+----------+
| count(*) |
+----------+
|       10 |
+----------+
1 row in set (0.00 sec)

-- status variables ARE incremented
mysql> show session status like '%handler%read%';
+-----------------------+-------+
| Variable_name         | Value |
+-----------------------+-------+
| Handler_read_first    | 1     |
| Handler_read_key      | 0     |
| Handler_read_last     | 0     |
| Handler_read_next     | 10    |
| Handler_read_prev     | 0     |
| Handler_read_rnd      | 0     |
| Handler_read_rnd_next | 0     |
+-----------------------+-------+
7 rows in set (0.02 sec)

Suggested fix:
Increment the status counters in InnoDB parallel count(*) implementation, since rows are actually read.

The 'Handler_read_first' status counter should be incremented when initializing the parallel scan context.

The 'Handler_read_next' status counter should be incremented here:
storage/innobase/row/row0pread.cc:583

I suppose the row should be listed as being traversed even if the row is not visible, so the Handler_read_next counter should actually be incremented before
if(!skip) {
...
}
in row0pread.cc

Hello Justin,

Thank you for the report.

regards,
Umesh

this also affects 8.0.0-dmr-enterprise-commercial-advanced
so I don't think it's to do with parallel count(*)