我最近将MySQL 5.1升级到5.7.8rc。
我在分析期间有“发送数据”状态的唯一问题。每个复杂或联合查询需要比预期更多的时间（对于简单查询也需要一些时间）。我试过最好的优化和原始查询。用Google搜索并尝试了所有可能的配置，但没有运气。
查询在MySQL 5.1中快速执行超级，但在5.7中没有。
还尝试了表格优化，分析，修复等。
Status Duration
starting 0.000515
checking permissions 0.000023
checking permissions 0.000016
checking permissions 0.000014
checking permissions 0.000014
checking permissions 0.000014
checking permissions 0.000014
checking permissions 0.000014
checking permissions 0.000014
checking permissions 0.000016
checking permissions 0.000014
checking permissions 0.000019
Opening tables 0.000079
init 0.000325
System lock 0.000068
optimizing 0.000079
statistics 0.001888
preparing 0.000151
Creating tmp table 0.000128
Sorting result 0.000027
optimizing 0.000034
statistics 0.000064
preparing 0.000047
Creating tmp table 0.000047
Sorting for group 0.000030
optimizing 0.000018
statistics 0.000022
preparing 0.000022
Creating tmp table 0.000043
Sorting for order 0.000023
executing 0.000016
Sending data 4.015596
Creating sort index 0.004766
executing 0.000010
Sending data 0.000159
executing 0.000008
Sending data 0.000025
Creating sort index 0.000349
end 0.000010
query end 0.000024
removing tmp table 0.000013
query end 0.000011
removing tmp table 0.000008
query end 0.000009
removing tmp table 0.000011
query end 0.000009
removing tmp table 0.000007
query end 0.000010
removing tmp table 0.000008
query end 0.000006
closing tables 0.000026
freeing items 0.000039
removing tmp table 0.000009
freeing items 0.000067
logging slow query 0.000047
cleaning up 0.000042
+----+--------------+-------------+------------+--------+-----------------------------------------------------+--------------------+---------+--------------------------------------+------+----------+------------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra |
+----+--------------+-------------+------------+--------+-----------------------------------------------------+--------------------+---------+--------------------------------------+------+----------+------------------------------------+
| 1 | PRIMARY | TIGM_GRP | NULL | index | PRIMARY,FLD_PARENT_GROUP_ID | PRIMARY | 2 | NULL | 87 | 0.33 | Using where |
| 1 | PRIMARY | TIPLD | NULL | range | FLD_PRICE_LEVEL_ID | FLD_PRICE_LEVEL_ID | 3 | NULL | 26 | 10.00 | Using index condition; Using where |
| 1 | PRIMARY | TIPLM | NULL | eq_ref | PRIMARY | PRIMARY | 2 | TIPLD.FLD_PRICE_LEVEL_ID | 1 | 10.00 | Using where |
| 1 | PRIMARY | TIGL | NULL | ref | FLD_GROUP_ID,fld_item_id | FLD_GROUP_ID | 2 | TIGM_GRP.FLD_GROUP_ID | 404 | 100.00 | NULL |
| 1 | PRIMARY | TIM | NULL | eq_ref | PRIMARY,FLD_ITEM_TYPE,FLD_ADDON_ID,FLD_PRODUCT_TYPE | PRIMARY | 4 | TIGL.FLD_ITEM_ID | 1 | 50.00 | Using where |
| 1 | PRIMARY | TSIAM | NULL | eq_ref | PRIMARY | PRIMARY | 4 | TIM.FLD_ADDON_ID | 1 | 10.00 | Using where |
| 1 | PRIMARY | TSIARM | NULL | ref | FLD_ADDON_ID | FLD_ADDON_ID | 5 | TIM.FLD_ADDON_ID | 1 | 10.00 | Using index condition; Using where |
| 1 | PRIMARY | TIPD | NULL | ref | FLD_ITEM_ID,FLD_PRICE_LEVEL_ID | FLD_ITEM_ID | 2 | TIM.FLD_ITEM_ID | 3 | 1.35 | Using index condition; Using where |
| 1 | PRIMARY | TIGM_PARENT | NULL | eq_ref | PRIMARY | PRIMARY | 2 | TIGM_GRP.FLD_PARENT_GROUP_ID | 1 | 100.00 | Using index |
| 2 | UNION | TIGM_GRP | NULL | index | PRIMARY,FLD_PARENT_GROUP_ID | PRIMARY | 2 | NULL | 87 | 0.06 | Using where |
| 2 | UNION | TIGM_PARENT | NULL | eq_ref | PRIMARY | PRIMARY | 2 | TIGM_GRP.FLD_PARENT_GROUP_ID | 1 | 100.00 | Using index |
|NULL| UNION RESULT | <union1,2> | NULL | ALL | NULL | NULL | NULL | NULL | NULL | NULL | Using temporary; Using filesort |
+----+--------------+-------------+------------+--------+-----------------------------------------------------+--------------------+---------+--------------------------------------+------+----------+------------------------------------+
[mysql]
# CLIENT #######################################################################
port = 3306
socket = /var/lib/mysql/mysql.sock
[mysqld]
# GENERAL ######################################################################
user = mysql
port = 3306
socket = /var/lib/mysql/mysql.sock
server_id = 32108
default_storage_engine = InnoDB
pid_file = /var/run/mysqld/mysqld.pid
optimizer_prune_level = 0
optimizer_search_depth = 0
max_length_for_sort_data = 8388608 #New
net_buffer_length = 1048576 #New
back_log = 80
symbolic-links = 0
log_bin_trust_function_creators = 1
net_read_timeout = 10 #90
net_write_timeout = 10 #120
net_retry_count = 30
thread_stack = 512K #192K
long_query_time = 10
tmpdir = /tmp
# MyISAM #######################################################################
key_buffer_size = 64M
read_buffer_size = 32M
read_rnd_buffer_size = 32M
bulk_insert_buffer_size = 16M
myisam_sort_buffer_size = 128M
myisam_max_sort_file_size = 1G
myisam_repair_threads = 1
memlock
max_allowed_packet = 32M
max_connect_errors = 100
sql_mode = STRICT_TRANS_TABLES,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION
sysdate-is-now = 1
explicit_defaults_for_timestamp = 1
innodb = FORCE
# Password policy disabled as per communication 05-Sep2017
validate_password = OFF
# DATA STORAGE ##################################################################
datadir = /var/lib/mysql/
# BINARY LOGGING ################################################################
log-bin = /var/lib/mysql/mysql-bin
expire-logs-days = 14
sync-binlog = 1
# REPLICATION ###################################################################
skip-slave-start = 1
relay-log = /var/log/mysql-realy-logs/relay-bin
slave-skip-errors = 1062 #,1053,1032,1237,1146
slave-net-timeout = 60
relay_log_purge = 1
# CACHES AND LIMITS #############################################################
tmp-table-size = 256M
max-heap-table-size = 256M
query_cache_min_res_unit = 12288 #8192 #New
query-cache-type = 1
query-cache-size = 32M #102400 #0 #256M
max-connections = 150
thread-cache-size = 10 #-1 #Auto resized
open-files-limit = 65535
table_definition_cache = 2000
table_open_cache = 4096 #3092
table_open_cache_instances = 4
sort_buffer_size = 128M
join_buffer_size = 512M
binlog_cache_size = 16M
query_cache_limit = 4M
# INNODB ########################################################################
innodb_fast_shutdown = 1
innodb_flush_method = O_DIRECT
innodb_log_group_home_dir = /mysql_redo_logs/mysql_redo_logs
innodb_log_files_in_group = 2
innodb_log_file_size = 1G
innodb_flush_log_at_trx_commit = 2
innodb_file_per_table = 1 #ON
innodb_buffer_pool_size = 32G #20G
innodb_buffer_pool_instances = 32
innodb_log_buffer_size = 64M
innodb_adaptive_hash_index = 1 #ON
innodb_thread_concurrency = 300 # "0" is default and means infinite (as and when needed). #250 #32
innodb_thread_sleep_delay = 1
innodb_flush_neighbors = 1
innodb_sync_array_size = 832 # Default is 768
skip-innodb_doublewrite #New
innodb_page_cleaners = 32 # Must be =innodb_buffer_pool_instances
innodb_sort_buffer_size = 512M
innodb_read_io_threads = 64
innodb_write_io_threads = 16
#innodb_concurrency_tickets = 429496729
innodb_max_dirty_pages_pct = 90
innodb_lock_wait_timeout = 10 #80
innodb_compression_level = 0 #New
innodb_lru_scan_depth = 512 #1024 #Default
# LOGGING #######################################################################
log_error = /var/lib/mysql/mysql-error.log
log_queries_not_using_indexes = 1
slow_query_log = 1
log_error_verbosity = 3
[mysqld_safe]
open-files-limit = 4096
malloc-lib = /usr/lib64/libtcmalloc_minimal.so.4
[mysqldump]
quick
max_allowed_packet = 64M
到目前为止，我已经尝试了大多数选项，目前我在缓存模式下启动了服务以便快速响应。
你能否帮我解决这个“发送数据”延迟问题
答案 0 :(得分：0)
5.1和5.7之间发生了很多优化程序更改。首先，ICP是新的;我在EXPLAIN中多次看到它。我们需要查看查询以进一步讨论。如果可能，请从5.1获取EXPLAIN。
同时，这里有一些关于my.cnf的小评论：
thread_stack = 512K #192K
通常，增加此设置无用。
optimizer_prune_level = 0
optimizer_search_depth = 0
是什么促使您设置为0？
innodb = FORCE
在5.7.5中弃用;建议在它成为错误之前删除。
slave-skip-errors = 1062 #,1053,1032,1237,1146
在地毯下清扫小鬼 - 你会在脚趾上留下脚趾。
innodb_buffer_pool_size = 32G #20G
除非你在同一台机器上有一些大型应用程序，否则44G可能会好一些。
innodb_page_cleaners = 32 # Must be =innodb_buffer_pool_instances
重新评论 - 从技术上讲，它自动限制为pool_instances;见https://dev.mysql.com/doc/refman/8.0/en/innodb-parameters.html#sysvar_innodb_page_cleaners（谢谢，WilsonHauck）
log_queries_not_using_indexes = 1
在我看来，这会使慢日志变得混乱。有趣的条目是那些超过long_query_time的条目。你在慢速日志中有任何有趣的疑问吗？
答案 1 :(得分：0)
大家 我想我弄明白了这个问题。 1）5.6生成的优化引擎和计划是元数据和用户查询（非关系连接）的结果混合。引擎信任用户并获得最佳计划。 但5.7不信任表之间的软关系，尽管它们被索引。 2）5.7期望明确的数据完整性，不要相信用户。
<强> POC： 我创建了与硬FK引用相关的相同表格， wala！发现了一个神奇的回应。没有FK执行的同一查询占用6.085秒（发送数据）， 并且强制执行数据完整性的查询花费了0.05秒（索引没有变化）。
所以我认为5.7明确要求强大的数据完整性约束。