在上一篇《内存分配统计视图 | 全方位认识 sys 系统库》中,我们介绍了sys 系统库如何查询内存事件统计信息和buffer pool统计信息,本期的内容先给大家介绍按照等待事件统计相关的视图(注意不要和《按 file 分组统计视图|全方位认识 sys 系统库》介绍的内容搞混了,这篇中介绍的等待事件仅针对文件IO等待事件,而本篇介绍的是所有的等待事件)。下面请跟随我们一起开始 sys 系统库的系统学习之旅吧~
01
wait_classes_global_by_avg_latency,x$wait_classes_global_by_avg_latency
按照事件大类(等待事件名称层级中前三层组件组成的名称前缀)分组(如:wait/io/table、wait/io/file、wait/lock/table)的等待事件平均延迟时间(总IO延迟时间/总IOS)等统计信息,默认按照平均延迟时间(执行时间)降序排序。数据来源:events_waits_summary_global_by_event_name
该视图会忽略空闲等待事件(idle事件)信息
下面我们看看使用该视图查询返回的结果。
# 不带x$前缀的视图
admin@localhost : sys 12:58:11> select * from wait_classes_global_by_avg_latency limit 3;
+--------------------+-------+---------------+-------------+-------------+-------------+
| event_class | total | total_latency | min_latency | avg_latency | max_latency |
+--------------------+-------+---------------+-------------+-------------+-------------+
| wait/lock/metadata | 2 | 56.57 m | 12.94 m | 28.28 m | 43.63 m |
| wait/synch/cond | 7980 | 4.37 h | 0 ps | 1.97 s | 5.01 s |
| wait/io/socket | 28988 | 21.02 s | 0 ps | 725.29 us | 103.18 ms |
+--------------------+-------+---------------+-------------+-------------+-------------+
3 rows in set (0.05 sec)
# 带x$前缀的视图
admin@localhost : sys 12:58:22> select * from x$wait_classes_global_by_avg_latency limit 3;
+--------------------+-------+-------------------+-----------------+-----------------------+------------------+
| event_class | total | total_latency | min_latency | avg_latency | max_latency |
+--------------------+-------+-------------------+-----------------+-----------------------+------------------+
| wait/lock/metadata | 2 | 3393932470401750 | 776378395041375 | 1696966235200875.0000 | 2617554075360375 |
| wait/synch/cond | 7980 | 15739342570225500 | 0 | 1972348693010.7143 | 5006888904375 |
| wait/io/socket | 28990 | 21024710924250 | 0 | 725240114.6689 | 103181011500 |
+--------------------+-------+-------------------+-----------------+-----------------------+------------------+
3 rows in set (0.02 sec)
视图字段含义如下:
event_class:事件类别,事件名称层级中前三层组件组成的名称前缀,如'wait/io/file/sql/slow_log',截取后保留'wait/io/file' 字符串作为事件类别
total:对应事件大类的事件总次数
total_latency:对应事件大类的事件总延迟时间(执行时间)
min_latency:对应事件大类的单次事件最小延迟时间(执行时间)
avg_latency:对应事件大类中,每个事件的平均延迟时间(执行时间)
max_latency:对应事件大类的单次事件在最大延迟时间(执行时间)
02
wait_classes_global_by_latency,x$wait_classes_global_by_latency
按照事件大类(等待事件名称前三层前缀)分组(如:wait/io/table、wait/io/file、wait/lock/table)的等待事件平均延迟时间等统计信息,默认情况下按照总延迟时间(执行时间)降序排序。数据来源:events_waits_summary_global_by_event_name
下面我们看看使用该视图查询返回的结果。
# 不带x$前缀的视图
admin@localhost : sys 12:58:26> select * from wait_classes_global_by_latency limit 3;
+--------------------+----------+---------------+-------------+-------------+-------------+
| event_class | total | total_latency | min_latency | avg_latency | max_latency |
+--------------------+----------+---------------+-------------+-------------+-------------+
| wait/synch/cond | 7983 | 4.38 h | 0 ps | 1.97 s | 5.01 s |
| wait/lock/metadata | 2 | 56.57 m | 12.94 m | 28.28 m | 43.63 m |
| wait/io/table | 16096791 | 4.59 m | 12.03 us | 17.11 us | 2.02 m |
+--------------------+----------+---------------+-------------+-------------+-------------+
3 rows in set (0.02 sec)
# 带x$前缀的视图
admin@localhost : sys 12:58:40> select * from x$wait_classes_global_by_latency limit 3;
+--------------------+----------+-------------------+-----------------+-----------------------+------------------+
| event_class | total | total_latency | min_latency | avg_latency | max_latency |
+--------------------+----------+-------------------+-----------------+-----------------------+------------------+
| wait/synch/cond | 7984 | 15759344050722375 | 0 | 1973865737815.9287 | 5006888904375 |
| wait/lock/metadata | 2 | 3393932470401750 | 776378395041375 | 1696966235200875.0000 | 2617554075360375 |
| wait/io/table | 16096791 | 275441586767625 | 12026625 | 17111583.7168 | 121243803313125 |
+--------------------+----------+-------------------+-----------------+-----------------------+------------------+
3 rows in set (0.02 sec)
视图字段含义如下:
该视图字段含义和wait_classes_global_by_avg_latency,x$wait_classes_global_by_avg_latency 视图字段含义相同,只是排序字段不同而已
03
waits_by_host_by_latency,x$waits_by_host_by_latency
按照主机和事件名称分组的等待事件统计信息,默认情况下按照主机名和总的等待事件延迟时间降序排序,数据来源:events_waits_summary_by_host_by_event_name
该视图忽略空闲等待事件(idle事件)信息
下面我们看看使用该视图查询返回的结果。
# 不带x$前缀的视图
admin@localhost : sys 12:58:43> select * from waits_by_host_by_latency limit 3;
+-------------+----------------------------------------------+-------+---------------+-------------+-------------+
| host | event | total | total_latency | avg_latency | max_latency |
+-------------+----------------------------------------------+-------+---------------+-------------+-------------+
| 10.10.20.14 | wait/io/socket/sql/client_connection | 24568 | 20.53 s | 835.48 us | 70.46 ms |
| 10.10.20.14 | wait/synch/mutex/innodb/trx_pool_mutex | 2326 | 14.59 s | 6.27 ms | 215.63 ms |
| 10.10.20.14 | wait/synch/cond/sql/MYSQL_BIN_LOG::COND_done | 1707 | 13.74 s | 8.05 ms | 43.33 ms |
+-------------+----------------------------------------------+-------+---------------+-------------+-------------+
3 rows in set (0.00 sec)
# 带x$前缀的视图
admin@localhost : sys 12:59:04> select * from x$waits_by_host_by_latency limit 3;
+-------------+----------------------------------------------+-------+----------------+-------------+--------------+
| host | event | total | total_latency | avg_latency | max_latency |
+-------------+----------------------------------------------+-------+----------------+-------------+--------------+
| 10.10.20.14 | wait/io/socket/sql/client_connection | 24568 | 20526083640375 | 835480125 | 70457480625 |
| 10.10.20.14 | wait/synch/mutex/innodb/trx_pool_mutex | 2326 | 14586650782125 | 6271131000 | 215632752375 |
| 10.10.20.14 | wait/synch/cond/sql/MYSQL_BIN_LOG::COND_done | 1707 | 13737760876125 | 8047897125 | 43332152250 |
+-------------+----------------------------------------------+-------+----------------+-------------+--------------+
3 rows in set (0.01 sec)
视图字段含义如下:
host:发起连接的主机名
event:等待事件名称
total:对应主机发生的等待事件总次数
total_latency:对应主机的等待事件总延迟时间
avg_latency:对应主机的等待事件的平均延迟时间
max_latency:对应主机的单次等待事件的最大延迟时间
04
waits_by_user_by_latency,x$waits_by_user_by_latency
按照用户和事件名称分组的等待事件统计信息,默认情况下按照用户名和总的等待事件延迟事件降序排序,数据来源:events_waits_summary_by_user_by_event_name
该视图忽略空闲等待事件(idle事件)信息
下面我们看看使用该视图查询返回的结果。
# 不带x$前缀的视图
admin@localhost : sys 12:59:07> select * from waits_by_user_by_latency limit 3;
+-------+---------------------------------------------------+----------+---------------+-------------+-------------+
| user | event | total | total_latency | avg_latency | max_latency |
+-------+---------------------------------------------------+----------+---------------+-------------+-------------+
| admin | wait/lock/metadata/sql/mdl | 2 | 56.57 m | 28.28 m | 43.63 m |
| admin | wait/synch/cond/sql/MDL_context::COND_wait_status | 3395 | 56.56 m | 999.66 ms | 1.00 s |
| admin | wait/io/table/sql/handler | 16096791 | 4.59 m | 17.11 us | 2.02 m |
+-------+---------------------------------------------------+----------+---------------+-------------+-------------+
3 rows in set (0.01 sec)
# 带x$前缀的视图
admin@localhost : sys 12:59:22> select * from x$waits_by_user_by_latency limit 3;
+-------+---------------------------------------------------+----------+------------------+------------------+------------------+
| user | event | total | total_latency | avg_latency | max_latency |
+-------+---------------------------------------------------+----------+------------------+------------------+------------------+
| admin | wait/lock/metadata/sql/mdl | 2 | 3393932470401750 | 1696966235200875 | 2617554075360375 |
| admin | wait/synch/cond/sql/MDL_context::COND_wait_status | 3395 | 3393839154564375 | 999658071750 | 1004173431750 |
| admin | wait/io/table/sql/handler | 16096791 | 275441586767625 | 17111250 | 121243803313125 |
+-------+---------------------------------------------------+----------+------------------+------------------+------------------+
3 rows in set (0.01 sec)
视图字段含义如下:
user:与该连接关联的用户名
其他字段与waits_by_host_by_latency,x$waits_by_host_by_latency 视图字段含义相同,不同的是waits_by_user_by_latency,x$waits_by_user_by_latency视图是按照用户名和事件名称分组
05
waits_global_by_latency,x$waits_global_by_latency
按照事件名称分组的等待事件统计信息,默认按照等待事件总延迟时间降序排序。数据来源:events_waits_summary_global_by_event_name
该视图忽略空闲等待事件(idle事件)信息
下面我们看看使用该视图查询返回的结果。
# 不带x$前缀的视图
admin@localhost : sys 12:59:25> select * from waits_global_by_latency limit 3;
+---------------------------------------------------+-------+---------------+-------------+-------------+
| events | total | total_latency | avg_latency | max_latency |
+---------------------------------------------------+-------+---------------+-------------+-------------+
| wait/synch/cond/sql/MYSQL_BIN_LOG::update_cond | 2891 | 3.45 h | 4.29 s | 5.01 s |
| wait/lock/metadata/sql/mdl | 2 | 56.57 m | 28.28 m | 43.63 m |
| wait/synch/cond/sql/MDL_context::COND_wait_status | 3395 | 56.56 m | 999.66 ms | 1.00 s |
+---------------------------------------------------+-------+---------------+-------------+-------------+
3 rows in set (0.02 sec)
# 带x$前缀的视图
admin@localhost : sys 12:59:40> select * from x$waits_global_by_latency limit 3;
+---------------------------------------------------+-------+-------------------+------------------+------------------+
| events | total | total_latency | avg_latency | max_latency |
+---------------------------------------------------+-------+-------------------+------------------+------------------+
| wait/synch/cond/sql/MYSQL_BIN_LOG::update_cond | 2892 | 12411771548807250 | 4291760563125 | 5006888904375 |
| wait/lock/metadata/sql/mdl | 2 | 3393932470401750 | 1696966235200875 | 2617554075360375 |
| wait/synch/cond/sql/MDL_context::COND_wait_status | 3395 | 3393839154564375 | 999658071750 | 1004173431750 |
+---------------------------------------------------+-------+-------------------+------------------+------------------+
3 rows in set (0.02 sec)
视图字段含义如下:
events:等待事件名称
其他字段含义和waits_by_host_by_latency,x$waits_by_host_by_latency 视图字段含义相同,不同的是waits_global_by_latency,x$waits_global_by_latency视图只按照事件名称分组
本期内容就介绍到这里,本期内容参考链接如下:
https://dev.mysql.com/doc/refman/5.7/en/sys-waits-global-by-latency.html
https://dev.mysql.com/doc/refman/5.7/en/sys-wait-classes-global-by-latency.html
https://dev.mysql.com/doc/refman/5.7/en/sys-wait-classes-global-by-avg-latency.html
https://dev.mysql.com/doc/refman/5.7/en/sys-waits-by-host-by-latency.html
https://dev.mysql.com/doc/refman/5.7/en/sys-waits-by-user-by-latency.html
"翻过这座山,你就可以看到一片海!"。坚持阅读我们的"全方位认识 sys 系统库"系列文章分享,你就可以系统地学完它。谢谢你的阅读,我们下期不见不散!
| 作者简介
罗小波·数据库技术专家
《千金良方——MySQL性能优化金字塔法则》、《数据生态:MySQL复制技术与生产实践》作者之一。熟悉MySQL体系结构,擅长数据库的整体调优,喜好专研开源技术,并热衷于开源技术的推广,在线上线下做过多次公开的数据库专题分享,发表过近100篇数据库相关的研究文章。
全文完。
Enjoy MySQL :)
叶老师的「MySQL核心优化」大课已升级到MySQL 8.0,扫码开启MySQL 8.0修行之旅吧