MySQL 8.0 information_schema系统库的改进

目录

• 测试5.7和8.0不同版本访问i_s库的性能

---

information_schema有何用？

• 通过i_s获取mysql的一些元数据信息
• 获取表的数据文件、索引文件的大小、碎片情况、表行数、自增列增长情况等
• 获取正在运行的事务有那些，是否有阻塞等
• 获取当前mysql的连接processlist等等

mysql8.0 之前的查询方式

会在查询information_schema 某个表时创建临时表

• 来自文件的元数据，扫描文件系统获取frm文件的表定义
• 存储引擎的详细信息，例如动态表统计信息
• 来自mysql服务器中全局数据结构的数据

在表的数量很多时，每次查询i_s会从文件系统中读取每个单独的frm文件，使用更多的cpu周期来打开表并准备相关的内存数据结构

mysql8.0 开始的查询方式

• 引入了基于innodb的本地数据字典表
• 表中不在有frm表定义文件
• 所有数据库表的元数据都存储在事务数据字典表中
• i_s中表被设计为数据字典表上的view（有些还是临时表0_0）

消除了以下成本
查询information_schema时创建的临时表
扫描文件系统目录以查找frm文件

改进
利用mysql优化器的全部功能，使用数据字典表上的索引来更好的查询

mysql5.7中表文件
ll test*
jul 10 10:52 testse.frm
jul 10 10:52 testse.ibd

mysql8.0中表文件

ll test*
jul 10 10:25 testse.ibd

mysql5.7 和mysql8.0 i_s 中tables 表的存在方式

mysql5.7.22 
show create table information_schema.tables\g
*************************** 1. row ***************************
       table: tables
create table: create temporary table `tables` (
  `table_catalog` varchar(512) not null default '',
  `table_schema` varchar(64) not null default '',
  `table_name` varchar(64) not null default '',
  `table_type` varchar(64) not null default '',
  `engine` varchar(64) default null,

mysql8.0.15
  show create table information_schema.tables\g
*************************** 1. row ***************************
                view: tables
         create view: create algorithm=undefined definer=`mysql.infoschema`@`localhost` sql security definer view `

• mysql5.7 中i_s中tables表是以临时表的形式存在的（查询该表就会创建临时表，创建的临时表过多，可能会导致mysql占用的内存暴涨，出现oom）
• mysql8.0 中i_s中tables表以视图的形式存在（查询该视图，不会创建临时表，会使用到视图中表的索引）

mysql5.7中获取表大小情况

select   table_name,   concat(format(data_length / 1024 / 1024, 2),   'm') as dbdata_size,   concat(format(index_length / 1024 / 1024, 2),   'm') as dbindex_size,   concat(     format((data_length + index_length) / 1024 / 1024 / 1024,2),   'g') as `db_size(g)`,   avg_row_length,   table_rows,   update_time from   information_schema.tables where table_schema = 'test'   and table_name = 'testse';
+------------+-------------+--------------+------------+----------------+------------+---------------------+
| table_name | dbdata_size | dbindex_size | db_size(g) | avg_row_length | table_rows | update_time         |
+------------+-------------+--------------+------------+----------------+------------+---------------------+
| testse     | 0.02m       | 0.02m        | 0.00g      |            862 |         19 | 2019-07-10 10:52:02 |
+------------+-------------+--------------+------------+----------------+------------+---------------------+


执行计划中出现了where，未用到索引（存储引擎检索数据后，server层进行过滤）
desc select   table_name,   concat(format(data_length / 1024 / 1024, 2),   'm') as dbdata_size,   concat(format(index_length / 1024 / 1024, 2),   'm') as dbindex_size,   concat(     format(data_length + index_length / 1024 / 1024 / 1024,     2   ),   'g') as `db_size(g)`,   avg_row_length,   table_rows,   update_time from   information_schema.tables where table_schema = 'test'   and table_name = 'testse';
+----+-------------+--------+------------+------+---------------+-------------------------+---------+------+------+----------+---------------------------------------------------+
| id | select_type | table  | partitions | type | possible_keys | key                     | key_len | ref  | rows | filtered | extra                                             |
+----+-------------+--------+------------+------+---------------+-------------------------+---------+------+------+----------+---------------------------------------------------+
|  1 | simple      | tables | null       | all  | null          | table_schema,table_name | null    | null | null |     null | using where; open_full_table; scanned 0 databases |
+----+-------------+--------+------------+------+---------------+-------------------------+---------+------+------+----------+---------------------------------------------------+

mysql8.0中获取表大小情况

select   table_name,   concat(format(data_length / 1024 / 1024, 2),   'm') as dbdata_size,   concat(format(index_length / 1024 / 1024, 2),   'm') as dbindex_size,   concat(     format((data_length + index_length) / 1024 / 1024 / 1024,2),   'g') as `db_size(g)`,   avg_row_length,   table_rows,   update_time from   information_schema.tables where table_schema = 'test'   and table_name = 'testse';
+------------+-------------+--------------+------------+----------------+------------+---------------------+
| table_name | dbdata_size | dbindex_size | db_size(g) | avg_row_length | table_rows | update_time         |
+------------+-------------+--------------+------------+----------------+------------+---------------------+
| testse     | 0.02m       | 0.02m        | 0.00g      |            862 |         19 | 2019-07-10 10:25:16 |
+------------+-------------+--------------+------------+----------------+------------+---------------------+



能使用到数据字典表的索引
desc select   table_name,   concat(format(data_length / 1024 / 1024, 2),   'm') as dbdata_size,   concat(format(index_length / 1024 / 1024, 2),   'm') as dbindex_size,   concat(     format((data_length + index_length) / 1024 / 1024 / 1024,2),   'g') as `db_size(g)`,   avg_row_length,   table_rows,   update_time from   information_schema.tables where table_schema = 'test'   and table_name = 'testse';
+----+-------------+-------+------------+--------+--------------------+------------+---------+-------------------------+------+----------+-------------+
| id | select_type | table | partitions | type   | possible_keys      | key        | key_len | ref                     | rows | filtered | extra       |
+----+-------------+-------+------------+--------+--------------------+------------+---------+-------------------------+------+----------+-------------+
|  1 | simple      | cat   | null       | index  | primary            | name       | 194     | null                    |    1 |   100.00 | using index |
|  1 | simple      | sch   | null       | eq_ref | primary,catalog_id | catalog_id | 202     | mysql.cat.id,const      |    1 |   100.00 | using index |
|  1 | simple      | tbl   | null       | eq_ref | schema_id          | schema_id  | 202     | mysql.sch.id,const      |    1 |   100.00 | using where |
|  1 | simple      | stat  | null       | const  | primary            | primary    | 388     | const,const             |    1 |   100.00 | null        |
|  1 | simple      | ts    | null       | eq_ref | primary            | primary    | 8       | mysql.tbl.tablespace_id |    1 |   100.00 | null        |
|  1 | simple      | col   | null       | eq_ref | primary            | primary    | 8       | mysql.tbl.collation_id  |    1 |   100.00 | using index |
+----+-------------+-------+------------+--------+--------------------+------------+---------+-------------------------+------+----------+-------------+

测试5.7和8.0不同版本访问i_s库的性能

机器

cat /etc/redhat-release | xargs echo '版本 ' && dmidecode -s system-product-name | xargs echo '是否虚拟化 ' && cat /proc/cpuinfo |grep "processor"|wc -l | xargs echo 'cpu核数 '
版本  centos linux release 7.5.1804 (core)
是否虚拟化  kvm
cpu核数  4

1、分别在mysql5.7和mysql8.0 中创建5个库，每个库中30个表（共1500个表），每个表记录数为10000

user=admin
passwd=admin
port=57222
host=127.0.0.1

#创建5个库,分别在每个库中创建30个表
for i in {1..5};do

mysql -u$user -p$passwd -p$port -h$host<<eof
  create database if not exists test_${i};

eof

sysbench /usr/local/share/sysbench/oltp_read_write.lua --mysql_storage_engine=innodb  --table-size=10000 --tables=30 --mysql-db=test_${i} --mysql-user=$user --mysql-password=$passwd --mysql-port=$port  --mysql-host=$host --threads=8 --time=10 --report-interval=1 --events=0 --db-driver=mysql prepare
done;

2、自定义访问i_s.tables表的sysbench脚本
cat tests/mytest.lua

require("oltp_common")


function thread_init(thread_id)
 drv=sysbench.sql.driver()
 con=drv:connect()
end


local function get_rand_db()
   return sysbench.rand.uniform(1, 5)
end



function event(thread_id)
local vid1
local dbprefix


vid1=get_rand_db()
dbprefix = "test_"

-- 生成5个db中的随机一个db,如 test_1 ,test_5
-- vid2=string.format("'%s%s'",test_,vid1)
vid2="'" .. dbprefix .. vid1 .. "'"

con:query("select   table_name,   concat(format(data_length / 1024 / 1024, 2),   'm') as dbdata_size,   concat(format(index_length / 1024 / 1024, 2),   'm') as dbindex_size,   concat(     format((data_length + index_length) / 1024 / 1024 / 1024,2),   'g') as `db_size(g)`,   avg_row_length,   table_rows,   update_time from   information_schema.tables where table_schema =" .. vid2)

end

function thread_done()
 con:disconnect()
end

3、脚本通过sysbench测试访问i_s.tables的qps
cat test_i_s_access.sh

user=admin
passwd=admin
host=127.0.0.1
#输入mysql端口参数
port=$1
# port=8015
#输入sysbench测试时间参数多少秒
run_time=$2

function get_create_tmp_tables(){
mysql -u$user -p$passwd -p$port -h$host<<eof
select @@version as 'mysql version';

show global  status like '%tmp_table%';
eof
} 2>/dev/null

function begin_test(){
  cd /usr/local/share/sysbench
   sysbench ./tests/mytest.lua --mysql-db=test_1  --mysql-host=$host --mysql-port=$port --mysql-user=$user --mysql-password=$passwd  --threads=40  --report-interval=10 --rand-type=uniform --time=$run_time    run
}


service mysqld${port} restart
get_create_tmp_tables
begin_test
get_create_tmp_tables

4、mysql5.7和mysql8.0 主要配置

mysql5.7.22

5.7.22-log
innodb_buffer_pool_size 128m
innodb_log_buffer_size  64m
innodb_log_file_size    48m
binlog_format   row
innodb_flush_log_at_trx_commit  1
sync_binlog     1

mysql8.0.15

8.0.15
innodb_buffer_pool_size 128m
innodb_log_buffer_size  64m
innodb_log_file_size    48m
binlog_format   row
innodb_flush_log_at_trx_commit  1
sync_binlog     1

5、mysql5.7开启测试
bash test_i_s_acess.sh 57222 60

shutting down mysql.. success!
starting mysql.. success!
mysql version
5.7.22-log
variable_name   value
created_tmp_tables      1
sysbench 1.1.0 (using bundled luajit 2.1.0-beta3)

running the test with following options:
number of threads: 40
report intermediate results every 10 second(s)
initializing random number generator from current time


initializing worker threads...

threads started!

[ 10s ] thds: 40 tps: 4765.73 qps: 4765.73 (r/w/o: 4765.73/0.00/0.00) lat (ms,95%): 18.95 err/s: 0.00 reconn/s:                                                                                                              0.00
[ 20s ] thds: 40 tps: 5409.00 qps: 5409.00 (r/w/o: 5409.00/0.00/0.00) lat (ms,95%): 17.95 err/s: 0.00 reconn/s:                                                                                                              0.00
[ 30s ] thds: 40 tps: 5154.45 qps: 5154.45 (r/w/o: 5154.45/0.00/0.00) lat (ms,95%): 18.61 err/s: 0.00 reconn/s:                                                                                                              0.00
[ 40s ] thds: 40 tps: 5383.50 qps: 5383.50 (r/w/o: 5383.50/0.00/0.00) lat (ms,95%): 17.63 err/s: 0.00 reconn/s:                                                                                                              0.00
[ 50s ] thds: 40 tps: 5456.11 qps: 5456.11 (r/w/o: 5456.11/0.00/0.00) lat (ms,95%): 17.63 err/s: 0.00 reconn/s:                                                                                                              0.00
[ 60s ] thds: 40 tps: 5458.66 qps: 5458.66 (r/w/o: 5458.66/0.00/0.00) lat (ms,95%): 17.63 err/s: 0.00 reconn/s:                                                                                                              0.00
sql statistics:
    queries performed:
        read:                            316322
        write:                           0
        other:                           0
        total:                           316322
    transactions:                        316322 (5270.99 per sec.)
    queries:                             316322 (5270.99 per sec.)
    ignored errors:                      0      (0.00 per sec.)
    reconnects:                          0      (0.00 per sec.)

throughput:
    events/s (eps):                      5270.9892
    time elapsed:                        60.0119s
    total number of events:              316322

latency (ms):
         min:                                    0.42
         avg:                                    7.59
         max:                                  858.18
         95th percentile:                       17.95
         sum:                              2399738.58

threads fairness:
    events (avg/stddev):           7908.0500/53.41
    execution time (avg/stddev):   59.9935/0.00

mysql version
5.7.22-log
variable_name   value
created_tmp_tables      316327  

cpu使用率接近100%

• mysql5.7.22 40个并发线程 压力测试60秒，访问i_s.tables 平均qps为5k，并创建了31w个临时表!

6、mysql8.0开启测试

shutting down mysql.. success!
starting mysql.... success!
mysql version
8.0.15
variable_name   value
created_tmp_tables      1
sysbench 1.1.0 (using bundled luajit 2.1.0-beta3)

running the test with following options:
number of threads: 40
report intermediate results every 10 second(s)
initializing random number generator from current time


initializing worker threads...

threads started!

[ 10s ] thds: 40 tps: 1283.76 qps: 1283.76 (r/w/o: 1283.76/0.00/0.00) lat (ms,95%): 41.10 err/s: 0.00 reconn/s: 0.00
[ 20s ] thds: 40 tps: 1286.43 qps: 1286.43 (r/w/o: 1286.43/0.00/0.00) lat (ms,95%): 39.65 err/s: 0.00 reconn/s: 0.00
[ 30s ] thds: 40 tps: 1333.00 qps: 1333.00 (r/w/o: 1333.00/0.00/0.00) lat (ms,95%): 37.56 err/s: 0.00 reconn/s: 0.00
[ 40s ] thds: 40 tps: 1280.79 qps: 1280.79 (r/w/o: 1280.79/0.00/0.00) lat (ms,95%): 40.37 err/s: 0.00 reconn/s: 0.00
[ 50s ] thds: 40 tps: 1313.57 qps: 1313.57 (r/w/o: 1313.57/0.00/0.00) lat (ms,95%): 38.94 err/s: 0.00 reconn/s: 0.00
[ 60s ] thds: 40 tps: 1222.95 qps: 1222.95 (r/w/o: 1222.95/0.00/0.00) lat (ms,95%): 42.61 err/s: 0.00 reconn/s: 0.00
sql statistics:
    queries performed:
        read:                            77248
        write:                           0
        other:                           0
        total:                           77248
    transactions:                        77248  (1286.95 per sec.)
    queries:                             77248  (1286.95 per sec.)
    ignored errors:                      0      (0.00 per sec.)
    reconnects:                          0      (0.00 per sec.)

throughput:
    events/s (eps):                      1286.9486
    time elapsed:                        60.0242s
    total number of events:              77248

latency (ms):
         min:                                    1.88
         avg:                                   31.08
         max:                                  692.98
         95th percentile:                       40.37
         sum:                              2400505.33

threads fairness:
    events (avg/stddev):           1931.2000/51.17
    execution time (avg/stddev):   60.0126/0.00

mysql version
8.0.15
variable_name   value
created_tmp_tables      2

cpu使用率接近100%

• mysql8.0.15 40个并发线程，压力测试60秒，访问i_s.tables 平均qps为1.2k，并创建了1个临时表!

结论

• mysql8.0开始查询i_s中表不会再从文件系统(frm, trg，opt文件)和myisam、innodb系统表读取元数据信息（8.0开始元数据信息统一存放在数据字典表中）
• mysql8.0 访问i_s.tables不会创建临时表，这减少了内存暴涨的可能，但访问i_s.tables的qps大约是mysql5.7.22的1/5，访问速度没有mysql5.7.22的快
• mysql8.0 访问i_s.tables 用户空间cpu消耗没有mysql5.7的高（没有创建临时表的缘故吧），但系统空间cpu消耗高于mysql5.7！

参考
mysql 8.0: improvements to information_schema