这里只讨论原生DDL, 因为pt-os或者ghost的话,会自带进度的显示。
对于原生的DDL,目前有2种方法观测进度
在MySQL 5.7.6或者更高版本,能够通过 performance_schema 观察alter table的进度。方法如下:
UPDATE performance_schema.setup_instruments SET ENABLED = 'YES' WHERE NAME LIKE 'stage/innodb/alter%';
UPDATE performance_schema.setup_consumers SET ENABLED = 'YES' WHERE NAME LIKE '%stages%';
select EVENT_NAME,WORK_COMPLETED,WORK_ESTIMATED from performance_schema.events_stages_current;
返回的结果集有3列数据:
EVENT_NAME 执行到的阶段
WORK_COMPLETED 已执行的耗时
WORK_ESTIMATED 预估总的耗时
例子:
开一个会话,执行DDL:
alter table sbtest3 modify pad varchar(256) ;
另开一个会话,查看:
performance_schema> select EVENT_NAME,WORK_COMPLETED,WORK_ESTIMATED from performance_schema.events_stages_current;
+-------------------------------------------+----------------+----------------+
| EVENT_NAME | WORK_COMPLETED | WORK_ESTIMATED |
+-------------------------------------------+----------------+----------------+
| stage/sql/Waiting for table metadata lock | <null> | <null> |
| stage/sql/copy to tmp table | 21368 | 98712 |
+-------------------------------------------+----------------+----------------+
可以看出,当前进度: 21368/98712 = 21%
开一个会话,执行DDL
alter table sbtest9 change pad pad char(90)
另开一个会话,查看:
[(none)]> select * from sys.session where conn_id != connection_id() \G
*************************** 1. row ***************************
thd_id: 20666
conn_id: 20273
user: root@localhost
db: sbtest
command: Query
state: copy to tmp table
time: 34
current_statement: alter table sbtest9 change pad pad char(90)
execution_engine: PRIMARY
statement_latency: 34.44 s --> 可以看到当前的DDL已经执行的时长
progress: 10.53 --> 可以看到 进度10.53%
lock_latency: 13.00 us
cpu_latency: 0 ps
rows_examined: 0
rows_sent: 0
rows_affected: 0
tmp_tables: 0
tmp_disk_tables: 0
full_scan: NO
last_statement: NULL
last_statement_latency: NULL
current_memory: 1.24 MiB
last_wait: NULL
last_wait_latency: NULL
source: NULL
trx_latency: 34.44 s
trx_state: ACTIVE
trx_autocommit: NO
pid: 26299
program_name: mysql
原创声明:本文系作者授权腾讯云开发者社区发表,未经许可,不得转载。
如有侵权,请联系 cloudcommunity@tencent.com 删除。
原创声明:本文系作者授权腾讯云开发者社区发表,未经许可,不得转载。
如有侵权,请联系 cloudcommunity@tencent.com 删除。