【1】CPU與內存
（1）ps -ef|grep mysqld / top 查看mysql的 pid
（2）top -H -p mysql_pid
（3）mysql里對應 performance_schema.threads 查mysql線程與os線程對應關系
一、如何排查高 IO 問題？
作為一個DBA難免不了會遇到性能問題，那么我們遇到性能問題該如何進行排查呢？例如我們在高並發的業務下，出現業務響應慢，處理時間長我們又該如何入手進行排查，本篇文章將分析io高的情況下如何分析及定位。
二、環境復現
- 環境配置：本次測試使用128C_512G_4TSSD服務器配置，MySQL版本為8.0.27
- 場景模擬：使用sysbench創建5個表，每個表2億條數據，執行產生笛卡爾積查詢的sql語句，產生io,可以模擬業務壓力。首先使用sysbench進行數據壓測
三、系統層面底層故障排查
Shell> sysbench --test=/usr/local/share/sysbench/oltp_insert.lua --mysql-host=XXX --mysql-port=3306 --mysql-user=pcms --mysql-password=abc123 --mysql-db=sysbench --percentile=99 --table-size=2000000000 --tables=5 --threads=1000 prepare
使用sysbench進行模擬高並發
shell> sysbench --test=/usr/local/share/sysbench/oltp_write_only.lua --mysql-host=xxx --mysql-port=3306 --mysql-user=pcms --mysql-password=abc123 --mysql-db=sysbench --percentile=99 --table-size=2000000000 --tables=5 --threads=1000 --max-time=60000 --report-interval=1 --threads=1000 --max-requests=0 --mysql-ignore-errors=all run
執行笛卡爾積sql語句
mysql> select SQL_NO_CACHE b.id,a.k from sbtest_a a left join sbtest_b b on a.id=b.id group by a.k order by b.c desc;
3.1 檢查當前服務器狀態
shell> top
top - 19:49:05 up 10 days, 8:16, 2 users, load average: 72.56, 40.21, 17.08
Tasks: 1288 total, 1 running, 586 sleeping, 0 stopped, 0 zombie
%Cpu(s): 19.7 us, 4.2 sy, 0.0 ni, 75.9 id, 1.0 wa, 0.0 hi, 0.2 si, 0.0 st
KiB Mem : 53542118+total, 23667507+free, 22735366+used, 71392448 buff/cache
KiB Swap: 0 total, 0 free, 0 used. 23128006+avail Mem
由上可知：目前一分鍾負載為72.56，且呈上升趨勢，並且存在io壓力
3.2 查看當前各個磁盤設備的io情況（iostat）
shell> iostat -m -x 1
Linux 4.14.0-115.el7a.0.1.aarch64 (mysql-4) 01/08/2022 _aarch64_ (128 CPU)
Device: rrqm/s wrqm/s r/s w/s rMB/s wMB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sda 0.00 0.00 14213.00 27430.00 222.08 465.15 33.80 5.39 0.13 0.14 0.12 0.02 86.00
sdb 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
dm-0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
dm-1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
dm-2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
由上可知：目前有多塊物理磁盤，sda磁盤的io壓力較大
3.3 檢查sda磁盤當前的io讀寫情況（iostat -d /dev/sda）
shell> iostat -d /dev/sda -m -x 1
Linux 4.14.0-115.el7a.0.1.aarch64 (mysql-4) 01/08/2022 _aarch64_ (128 CPU)
Device: rrqm/s wrqm/s r/s w/s rMB/s wMB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sda 0.00 0.17 7.86 81.23 0.29 3.96 97.88 0.23 2.53 0.22 2.76 0.04 0.33
Device: rrqm/s wrqm/s r/s w/s rMB/s wMB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sda 0.00 0.00 21109.00 42839.00 329.81 710.90 33.33 19.47 0.30 0.16 0.37 0.02 96.00
Device: rrqm/s wrqm/s r/s w/s rMB/s wMB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sda 0.00 0.00 19287.00 41404.00 301.36 692.29 33.53 15.73 0.26 0.18 0.30 0.02 93.00
Device: rrqm/s wrqm/s r/s w/s rMB/s wMB/s avgrq-sz avgqu-sz await r_await w_await svctm %util
sda 0.00 0.00 22135.00 43044.00 345.86 1165.18 47.48 100.87 1.55 0.20 2.24 0.01 97.00
由上可知：目前sda磁盤的壓力比較大，每秒寫入比每秒讀差距較大，證明目前有大量的io寫入
3.4 檢查sda磁盤中哪個應用程序占用的io比較高（pidstat ）
shell> pidstat -d 1
Linux 4.14.0-115.el7a.0.1.aarch64 (mysql-4) 01/08/2022 _aarch64_ (128 CPU)
08:01:43 PM UID PID kB_rd/s kB_wr/s kB_ccwr/s Command
08:01:44 PM 1000 73739 62018.35 171346.79 0.00 mysqld
08:01:44 PM UID PID kB_rd/s kB_wr/s kB_ccwr/s Command
08:01:45 PM 1000 73739 145328.00 435216.00 0.00 mysqld
08:01:45 PM UID PID kB_rd/s kB_wr/s kB_ccwr/s Command
08:01:46 PM 1000 73739 141488.00 433232.00 0.00 mysqld
由上可知：占用io高的應用程序是mysql，且pid為73739
3.5 分析應用程序中哪一個線程占用的io比較高
shell> pidstat -dt -p 73739 1 執行兩三秒即可
Average: 1000 - 73823 0.00 233133.98 0.00 |__mysqld
Average: 1000 - 74674 0.00 174291.26 0.00 |__mysqld
11:56:18 PM 1000 - 74770 124928.00 74688.00 0.00 |__mysqld
11:56:17 PM 1000 - 74770 124603.77 73358.49 0.00 |__mysqld
Average: 1000 - 74770 124761.17 74003.88 0.00 |__mysqld
由上可知：74770這個線程占用的io比較高
3.6 分析這個線程在干什么？
Shell> perf trace -t 74770 -o /tmp/tmp_aa.pstrace
Shell> cat /tmp/tmp_aa.pstrace
2850.656 ( 1.915 ms): futex(uaddr: 0x653ae9c4, op: WAIT|PRIVATE_FLAG, val: 1) = 0
2852.572 ( 0.001 ms): futex(uaddr: 0x653ae990, op: WAKE|PRIVATE_FLAG, val: 1) = 0
2852.601 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f68) = 0
2852.690 ( 0.040 ms): write(fd: 159, buf: 0xd7a30020, count: 65536) = 65536
2852.796 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f68) = 0
2852.798 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f58) = 0
2852.939 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f38) = 0
2852.950 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f68) = 0
2852.977 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f68) = 0
2853.029 ( 0.035 ms): write(fd: 64, buf: 0xcd51e020, count: 65536) = 65536
2853.164 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f68) = 0
2853.167 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f58) = 0
2853.302 ( 0.001 ms): clock_gettime(which_clock: MONOTONIC, tp: 0xfff7bd470f38) = 0
由上可知：目前這個線程在寫入多個文件，fd為文件句柄，文件句柄號有64、159
3.7 查看這個文件句柄是什么
shell> lsof -p 73739|grep 159u
mysqld 73739 mysql 159u REG 8,0 212143246 7046482357 /mysql/mysqldata/16320fff-5fd5-4c47-889a-a9e1a8591d0d/tmp/#7046482357 (deleted)
[root@mysql-4 ~]# lsof -p 73739|grep 64u
mysqld 73739 mysql 64u REG 8,0 211872724 6979323031 /mysql/mysqldata/16320fff-5fd5-4c47-889a-a9e1a8591d0d/tmp/#6979323031 (deleted)
由上可知：這個線程在大量的寫入臨時文件
四、分析MySQL應用程序
4.1 查看當前的會話列表
mysql> select * from information_schema.processlist where command !='sleep';
| 9 | pcms | 172.16.76.12:57596 | sysbench | Query | 67 | executing | select SQL_NO_CACHE b.id,a.k from sbtest_a a left join sbtest_b b on a.id=b.id group by a.k order by b.c desc | 66477 | 0 | 0 |
由上可知：目前這個sql已經執行了67s，且此sql使用了group by和order by，必然會產生io
4.2 通過線程號查詢會話
mysql> select * from threads where thread_os_id=74770\G;
*************************** 1. row ***************************
THREAD_ID: 95
NAME: thread/sql/one_connection
TYPE: FOREGROUND
PROCESSLIST_ID: 9
PROCESSLIST_USER: pcms
PROCESSLIST_HOST: 172.16.76.12
PROCESSLIST_DB: sysbench
PROCESSLIST_COMMAND: Query
PROCESSLIST_TIME: 91
PROCESSLIST_STATE: NULL
PROCESSLIST_INFO: select SQL_NO_CACHE b.id,a.k from sbtest_a a left join sbtest_b b on a.id=b.id group by a.k order by b.c desc
PARENT_THREAD_ID: 1
ROLE: NULL
INSTRUMENTED: YES
HISTORY: YES
CONNECTION_TYPE: TCP/IP
THREAD_OS_ID: 74770
RESOURCE_GROUP: USR_default
1 row in set (0.00 sec)
由上可知：通過查詢threads表可以進行驗證，該線程在頻繁創建臨時表的原因就來源於此sql
4.3 查看該sql語句的執行計划，進行進一步認證
mysql> explain select SQL_NO_CACHE b.id,a.k from sbtest_a a left join sbtest_b b on a.id=b.id group by a.k order by b.c desc\G;
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: a
partitions: NULL
type: ALL
possible_keys: NULL
key: NULL
key_len: NULL
ref: NULL
rows: 1
filtered: 100.00
Extra: Using temporary; Using filesort
*************************** 2. row ***************************
id: 1
select_type: SIMPLE
table: b
partitions: NULL
type: eq_ref
possible_keys: PRIMARY
key: PRIMARY
key_len: 4
ref: sysbench.a.id
rows: 1
filtered: 100.00
Extra: NULL
2 rows in set, 2 warnings (0.00 sec)
由上可知：該sql的執行計划用到了臨時表及臨時文件，符合
4.4 查看全局狀態進一步進行確認
mysql> show global status like '%tmp%';
+-------------------------+-------+
| Variable_name | Value |
+-------------------------+-------+
| Created_tmp_disk_tables | 3 |
| Created_tmp_files | 22 |
| Created_tmp_tables | 8 |
+-------------------------+-------+
多執行幾次，可以看出tmp_files和tmp_disk_tables的值在增長，證明在大量的創建臨時文件及磁盤臨時表，符合該線程的行為
五、故障處理
通過上述的一系列排查，我們已經分析出來，目前sda磁盤的io使用率最高，且mysqld程序占用的最多，通過排查有一個線程在頻繁的創建臨時表或臨時文件且通過登錄mysql排查會話及線程視圖可以找到是由某一個慢sql導致的，查看此慢sql的執行計划也會創建臨時表和臨時文件符合我們之前排查的預期。此時我們就需要針對此慢sql進行優化，優化步驟由DBA進行處理，此處進行忽略。慢sql優化完成后可以進行io的繼續觀察，看io是否有下降
六、代碼分析
我們可以使用pstack進行跟蹤線程號，獲取當前的線程堆棧信息。切記pstack會調用gdb進行debug調試
shell> pstack 74770 >/tmp/74770.pstack
Thread 1 (process 74770):
#0 ha_innobase::general_fetch (this=0xea654228, buf=0xea662028 "\212t\317\030\002", direction=1, match_mode=0 ) at /builds/naiwei.fang/percona-server/storage/innobase/handler/ha_innodb.cc:11159
#1 0x0000000000d9913c in handler::ha_rnd_next (this=0xea654228, buf=0xea662028 "\212t\317\030\002") at /build
s/naiwei.fang/percona-server/sql/handler.cc:3173
#2 0x0000000000f77db0 in TableScanIterator::Read (this=0xd256d5e8) at /builds/naiwei.fang/percona-server/sql/
row_iterator.h:208
#3 0x000000000124c714 in WriteRowsToChunks (xxhash_seed=899339, write_to_build_chunk=true, write_rows_with_nu
ll_in_join_key=false, join_key_buffer=0xd01fdb98, tables_to_get_rowid_for=0, chunks=0xd01fdb58, join_condition
s=..., tables=..., iterator=0xd256d5e8, thd=0xdb888000) at /builds/naiwei.fang/percona-server/sql/hash_join_it
erator.cc:282
#4 HashJoinIterator::BuildHashTable (this=this@entry=0xd01fd028) at /builds/naiwei.fang/percona-server/sql/ha
sh_join_iterator.cc:495
#5 0x000000000124c8ac in Init (this=0xd01fd028) at /builds/naiwei.fang/percona-server/sql/hash_join_iterator.
cc:203
#6 HashJoinIterator::Init (this=0xd01fd028) at /builds/naiwei.fang/percona-server/sql/hash_join_iterator.cc:1
45
#7 0x00000000010eca14 in Query_expression::ExecuteIteratorQuery (this=0xdec3a8b8, thd=thd@entry=0xdb888000) a
t /builds/naiwei.fang/percona-server/sql/sql_union.cc:1224
#8 0x00000000010ecccc in Query_expression::execute (this=this@entry=0xdec3a8b8, thd=thd@entry=0xdb888000) at
/builds/naiwei.fang/percona-server/sql/sql_union.cc:1284
#9 0x0000000001083db0 in Sql_cmd_dml::execute_inner (this=0xd256bcb0, thd=0xdb888000) at /builds/naiwei.fang/
percona-server/sql/sql_select.cc:791
#10 0x000000000108cac8 in Sql_cmd_dml::execute (this=0xd256bcb0, thd=0xdb888000) at /builds/naiwei.fang/percon
a-server/sql/sql_select.cc:575
#11 0x00000000010384e8 in mysql_execute_command (thd=thd@entry=0xdb888000, first_level=first_level@entry=true)
at /builds/naiwei.fang/percona-server/sql/sql_parse.cc:4677
#12 0x000000000103b314 in dispatch_sql_command (thd=thd@entry=0xdb888000, parser_state=parser_state@entry=0xff
f7bd4735b0, update_userstat=update_userstat@entry=false) at /builds/naiwei.fang/percona-server/sql/sql_parse.c
c:5273
#13 0x000000000103ccf0 in dispatch_command (thd=thd@entry=0xdb888000, com_data=0xffffb467c4d0, com_data@entry=
0xfff7bd474640, command=COM_QUERY) at /builds/naiwei.fang/percona-server/sql/sql_parse.cc:1938
#14 0x000000000103da40 in do_command (thd=thd@entry=0xdb888000) at /builds/naiwei.fang/percona-server/sql/sql_
parse.cc:1386
#15 0x0000000001152ca8 in handle_connection (arg=arg@entry=0xda53ab10) at /builds/naiwei.fang/percona-server/s
ql/conn_handler/connection_handler_per_thread.cc:307
#16 0x00000000022bc3ec in pfs_spawn_thread (arg=<optimized out>) at /builds/naiwei.fang/percona-server/storage
/perfschema/pfs.cc:2899
#17 0x0000ffffb43c7c48 in start_thread () from /lib64/libpthread.so.0
#18 0x0000ffffb3c0f600 in thread_start () from /lib64/libc.so.6
【參考文檔】
參考站姿：Linux 磁盤 io 利用率高排查：https://mp.weixin.qq.com/s?__biz=MzkzMTIzMDgwMg==&mid=2247487912&idx=1&sn=fcfe1eb1fa56fc769abd9517dd6fe30c&chksm=c26f7495f518fd83fb7b8e263d03750cad09c18051aaccd8174d1436b103123891f26540b9b0&scene=21#wechat_redirect