如何把mysqld压测到崩溃重启

小编给大家分享一下如何把mysqld压测到崩溃重启，希望大家阅读完这篇文章之后都有所收获，下面让我们一起去探讨吧！

一、压测环境工具准备：

centos7.5

sysbench2.0.9

mysql5.7.22

机器配置：宿主机是vmware esxi

DELL R730

硬盘：普通10K SAS

内存：18G

CPU：8核

非常普通的cpu：

[root@yw-gz-hd-test-211 log]# lscpu Architecture:          x86_64CPU op-mode(s):        32-bit, 64-bitByte Order:            Little EndianCPU(s):                8On-line CPU(s) list:   0-7Thread(s) per core:    1Core(s) per socket:    1Socket(s):             8NUMA node(s):          1Vendor ID:             GenuineIntelCPU family:            6Model:                 79Model name:            Intel(R) Xeon(R) CPU E5-2640 v4 @ 2.40GHzStepping:              1CPU MHz:               2399.361BogoMIPS:              4799.99Hypervisor vendor:     VMwareVirtualization type:   fullL1d cache:             32KL1i cache:             32KL2 cache:              256KL3 cache:              25600KNUMA node0 CPU(s):     0-7Flags:                 fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts mmx fxsr sse sse2 ss syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts nopl xtopology tsc_reliable nonstop_tsc eagerfpu pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 invpcid rtm rdseed adx smap xsaveopt arat

编译安装好mysql，设置 innodb_buffer_pool_size=5G innodb_buffer_pool_instance=5. 其他参数更改redo 为4组，io thread 为8 等等一些参数。

二、开始准备压测数据库：

插入10张表，每个表数据1000万，整个msyql库25G。

[root@yw-gz-hd-test-211 ~]# ls /data/mysql3308/sbtest/ -lhtotal 25G-rw-r----- 1 mysql mysql   61 Jul 17 19:24 db.opt-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest10.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest10.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest1.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest1.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest2.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest2.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest3.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest3.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest4.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest4.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest5.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest5.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest6.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest6.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest7.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest7.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest8.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest8.ibd-rw-r----- 1 mysql mysql 8.5K Jul 17 19:32 sbtest9.frm-rw-r----- 1 mysql mysql 2.5G Jul 18 14:58 sbtest9.ibd

[root@yw-gz-hd-test-211 ~]# ls /data/mysql3308/ -lhtotal 1.4G-rw-r----- 1 mysql mysql   56 Jul 17 17:56 auto.cnf-rw-r----- 1 mysql mysql 1.5K Jul 18 14:17 ib_buffer_pool-rw-r----- 1 mysql mysql 384M Jul 18 15:13 ibdata1-rw-r----- 1 mysql mysql 256M Jul 18 15:13 ib_logfile0-rw-r----- 1 mysql mysql 256M Jul 18 14:50 ib_logfile1-rw-r----- 1 mysql mysql 256M Jul 18 15:13 ib_logfile2-rw-r----- 1 mysql mysql 256M Jul 18 14:49 ib_logfile3-rw-r----- 1 mysql mysql  12M Jul 18 15:21 ibtmp1drwxr-x--- 2 mysql mysql 4.0K Jul 17 17:56 mysqlsrwxrwxrwx 1 mysql mysql    0 Jul 18 14:42 mysql.sock-rw------- 1 mysql mysql    6 Jul 18 14:42 mysql.sock.lockdrwxr-x--- 2 mysql mysql 8.0K Jul 17 17:56 performance_schemadrwxr-x--- 2 mysql mysql 4.0K Jul 17 19:36 sbtestdrwxr-x--- 2 mysql mysql 8.0K Jul 17 17:56 sys-rw-r----- 1 mysql mysql    6 Jul 18 14:42 yw-gz-hd-test-211.pid

三、开始压测：

首先300个线程，开始上。你会发现，立马报错：

FATAL: mysql_stmt_prepare() failedFATAL: MySQL error: 1461 "Can't create more than max_prepared_stmt_count statements (current value: 100000)"

百度一下，设置一下参数可以解决：max_prepared_stmt_count=150000

四、高潮出现：

错误排除，压测到线程300个，总共时长是240秒，等到压测到120秒的时候，mysql进程突然奔溃。错误日志中没有记录mysql奔溃的原因，只记录到mysql崩溃后，被mysqld_safe 进程监控，然后立即拉起mysqld 进程。mysqld_safe 进程会一直监控mysqld进程，发现死掉，立即拉起mysqld进程。我怀疑是内存不够。但是没有证据证明：是内存不够导致的mysqld进程奔溃。这个时候，我发现top命令还是很好用的。怎么用呢？让我娓娓道来。前面不是讲到了，压测刚开始的120秒，没有问题，你可以在这个压测0~120秒的时候，打开top，你观察mysqld线程使用内存情况。你观察RES这一列，你会发现，mysqld进程RES值，从500M一直增长，增长到5G的时候，duang~，崩溃了。看出来了吧，mysql也有承受不住的时候。

为了验证自己是猜想，很简单，不要更改任何参数，增加机器内存到18G。再一次压测，验证了我的想法，mysqld进程再300个并发线程中使用掉了6G内存。

来看下300个并发压测情况

[root@yw-gz-hd-test-211 ~]# sysbench /usr/share/sysbench/oltp_read_write.lua --db-driver=mysql --mysql-host=localhost --mysql-socket=/data/mysql3308/mysql.sock  --mysql-port=3308  --mysql-db=sbtest --mysql-user=root --mysql-password=123456 --table_size=10000000 --tables=10 --threads=300  --time=240 --report-interval=30 runsysbench 1.0.9 (using system LuaJIT 2.0.4)Running the test with following options:Number of threads: 300Report intermediate results every 30 second(s)Initializing random number generator from current timeInitializing worker threads...Threads started![ 30s ] thds: 300 tps: 189.42 qps: 3911.05 (r/w/o: 2753.06/769.16/388.83) lat (ms,95%): 3151.62 err/s: 0.00 reconn/s: 0.00[ 60s ] thds: 300 tps: 406.65 qps: 8146.63 (r/w/o: 5702.77/1630.57/813.30) lat (ms,95%): 1903.57 err/s: 0.00 reconn/s: 0.00[ 90s ] thds: 300 tps: 1027.51 qps: 20561.94 (r/w/o: 14391.74/4115.19/2055.01) lat (ms,95%): 909.80 err/s: 0.00 reconn/s: 0.00[ 120s ] thds: 300 tps: 915.33 qps: 18308.17 (r/w/o: 12818.23/3659.27/1830.67) lat (ms,95%): 802.05 err/s: 0.00 reconn/s: 0.00[ 150s ] thds: 300 tps: 848.33 qps: 16954.26 (r/w/o: 11865.99/3391.60/1696.67) lat (ms,95%): 787.74 err/s: 0.00 reconn/s: 0.00[ 180s ] thds: 300 tps: 1015.47 qps: 20327.15 (r/w/o: 14231.78/4064.44/2030.93) lat (ms,95%): 682.06 err/s: 0.00 reconn/s: 0.00[ 210s ] thds: 300 tps: 1293.73 qps: 25882.66 (r/w/o: 18120.80/5174.40/2587.47) lat (ms,95%): 493.24 err/s: 0.00 reconn/s: 0.00[ 240s ] thds: 300 tps: 1705.07 qps: 33979.32 (r/w/o: 23772.88/6803.07/3403.37) lat (ms,95%): 419.45 err/s: 0.00 reconn/s: 0.00SQL statistics:    queries performed:        read:                            3110016        write:                           888576        other:                           444288        total:                           4442880    transactions:                        222144 (924.53 per sec.)    queries:                             4442880 (18490.54 per sec.)    ignored errors:                      0      (0.00 per sec.)    reconnects:                          0      (0.00 per sec.)General statistics:    total time:                          240.2250s    total number of events:              222144Latency (ms):         min:                                  2.52         avg:                                324.16         max:                              50333.39         95th percentile:                   1050.76         sum:                            72010070.69Threads fairness:    events (avg/stddev):           740.4800/78.20    execution time (avg/stddev):   240.0336/0.06

成绩还不错，QPS：18490，TPS：924。95%的响应时间是1050ms，就是1秒，可以接受

来看看600并发连接线程情况

[root@yw-gz-hd-test-211 ~]# sysbench /usr/share/sysbench/oltp_read_write.lua --db-driver=mysql --mysql-host=localhost --mysql-socket=/data/mysql3308/mysql.sock  --mysql-port=3308  --mysql-db=sbtest --mysql-user=root --mysql-password=123456 --table_size=10000000 --tables=10 --threads=600  --time=240 --report-interval=30 runsysbench 1.0.9 (using system LuaJIT 2.0.4)Running the test with following options:Number of threads: 600Report intermediate results every 30 second(s)Initializing random number generator from current timeInitializing worker threads...Threads started![ 30s ] thds: 600 tps: 177.45 qps: 3866.55 (r/w/o: 2740.46/751.20/374.88) lat (ms,95%): 6594.16 err/s: 0.00 reconn/s: 0.00[ 60s ] thds: 600 tps: 508.61 qps: 10190.12 (r/w/o: 7130.15/2042.76/1017.21) lat (ms,95%): 2828.87 err/s: 0.00 reconn/s: 0.00[ 90s ] thds: 600 tps: 833.10 qps: 16581.88 (r/w/o: 11603.42/3312.26/1666.20) lat (ms,95%): 1506.29 err/s: 0.00 reconn/s: 0.00[ 120s ] thds: 600 tps: 712.40 qps: 14275.18 (r/w/o: 9994.28/2856.20/1424.70) lat (ms,95%): 1589.90 err/s: 0.00 reconn/s: 0.00[ 150s ] thds: 600 tps: 828.53 qps: 16595.37 (r/w/o: 11637.94/3300.27/1657.17) lat (ms,95%): 1280.93 err/s: 0.00 reconn/s: 0.00[ 180s ] thds: 600 tps: 1152.15 qps: 23046.54 (r/w/o: 16115.87/4626.50/2304.17) lat (ms,95%): 1032.01 err/s: 0.00 reconn/s: 0.00[ 210s ] thds: 600 tps: 1422.39 qps: 28470.31 (r/w/o: 19918.05/5707.53/2844.74) lat (ms,95%): 707.07 err/s: 0.00 reconn/s: 0.00[ 240s ] thds: 600 tps: 1874.42 qps: 37511.54 (r/w/o: 26257.48/7505.04/3749.01) lat (ms,95%): 601.29 err/s: 0.00 reconn/s: 0.00SQL statistics:    queries performed:        read:                            3161774        write:                           903364        other:                           451682        total:                           4516820    transactions:                        225841 (939.46 per sec.)    queries:                             4516820 (18789.23 per sec.)    ignored errors:                      0      (0.00 per sec.)    reconnects:                          0      (0.00 per sec.)General statistics:    total time:                          240.3923s    total number of events:              225841Latency (ms):         min:                                  2.75         avg:                                637.78         max:                              44200.42         95th percentile:                   1678.14         sum:                            144036928.60Threads fairness:    events (avg/stddev):           376.4017/48.51    execution time (avg/stddev):   240.0615/0.03

这个时候我们看到大量的慢查询语句，95%响应时间是1678ms，就是1.6秒，有些慢了。看看慢查询都是些什么语句：

# Time: 2018-07-18T14:22:07.662597+08:00# User@Host: root[root] @ localhost []  Id:   592# Query_time: 7.400737  Lock_time: 0.000028 Rows_sent: 0  Rows_examined: 1SET timestamp=1531894927;UPDATE sbtest5 SET k=k+1 WHERE id=5024619;# Time: 2018-07-18T14:22:07.662786+08:00# User@Host: root[root] @ localhost []  Id:   202# Query_time: 4.220504  Lock_time: 0.000027 Rows_sent: 0  Rows_examined: 1SET timestamp=1531894927;UPDATE sbtest5 SET k=k+1 WHERE id=5024572;# Time: 2018-07-18T14:22:07.662829+08:00# User@Host: root[root] @ localhost []  Id:   544# Query_time: 3.662601  Lock_time: 0.000021 Rows_sent: 0  Rows_examined: 1SET timestamp=1531894927;DELETE FROM sbtest5 WHERE id=5024577;# Time: 2018-07-18T14:22:07.662634+08:00# User@Host: root[root] @ localhost []  Id:   402# Query_time: 4.832428  Lock_time: 0.000023 Rows_sent: 0  Rows_examined: 1SET timestamp=1531894927;UPDATE sbtest5 SET c='53575816661-90198037463-61731021712-17992612508-02527517402-89815419518-53211578757-17129425245-97225103738-94879199437' WHERE id=5024586;

都是更新语句。这些语句非常耗费IO的

再来看看900个并发线程的情况。

[root@yw-gz-hd-test-211 ~]# sysbench /usr/share/sysbench/oltp_read_write.lua --db-driver=mysql --mysql-host=localhost --mysql-socket=/data/mysql3308/mysql.sock  --mysql-port=3308  --mysql-db=sbtest --mysql-user=root --mysql-password=123456 --table_size=10000000 --tables=10 --threads=900  --time=240 --report-interval=30 runsysbench 1.0.9 (using system LuaJIT 2.0.4)Running the test with following options:Number of threads: 900Report intermediate results every 30 second(s)Initializing random number generator from current timeInitializing worker threads...Threads started![ 30s ] thds: 900 tps: 347.86 qps: 7432.37 (r/w/o: 5273.60/1433.11/725.65) lat (ms,95%): 5124.81 err/s: 0.00 reconn/s: 0.00[ 60s ] thds: 900 tps: 561.28 qps: 11176.43 (r/w/o: 7801.59/2252.28/1122.55) lat (ms,95%): 10158.80 err/s: 0.00 reconn/s: 0.00[ 90s ] thds: 900 tps: 643.33 qps: 12944.09 (r/w/o: 9077.29/2580.13/1286.67) lat (ms,95%): 2932.60 err/s: 0.00 reconn/s: 0.00[ 120s ] thds: 900 tps: 360.53 qps: 7200.07 (r/w/o: 5039.67/1439.33/721.07) lat (ms,95%): 6135.91 err/s: 0.00 reconn/s: 0.00[ 150s ] thds: 900 tps: 728.53 qps: 14524.71 (r/w/o: 10134.68/2933.03/1457.00) lat (ms,95%): 2585.31 err/s: 0.00 reconn/s: 0.00[ 180s ] thds: 900 tps: 1268.27 qps: 25410.63 (r/w/o: 17798.37/5075.80/2536.47) lat (ms,95%): 1561.52 err/s: 0.00 reconn/s: 0.00[ 210s ] thds: 900 tps: 1676.04 qps: 33561.08 (r/w/o: 23477.06/6731.82/3352.21) lat (ms,95%): 1869.60 err/s: 0.00 reconn/s: 0.00[ 240s ] thds: 900 tps: 2290.01 qps: 45719.85 (r/w/o: 31996.75/9148.79/4574.31) lat (ms,95%): 1352.03 err/s: 0.00 reconn/s: 0.00SQL statistics:    queries performed:        read:                            3318098        write:                           948028        other:                           474014        total:                           4740140    transactions:                        237007 (985.74 per sec.)    queries:                             4740140 (19714.74 per sec.)    ignored errors:                      0      (0.00 per sec.)    reconnects:                          0      (0.00 per sec.)General statistics:    total time:                          240.4346s    total number of events:              237007Latency (ms):         min:                                  2.76         avg:                                911.43         max:                              31437.18         95th percentile:                   2778.39         sum:                            216015485.39Threads fairness:    events (avg/stddev):           263.3411/37.88    execution time (avg/stddev):   240.0172/0.02

看到了，95%响应时间是2.7秒，数据库mysql响应时间越来越慢，越来越不堪重负。崩溃就在一瞬间。如我所见，innodb_buffer_pool_size=18G时候，1000个并发线程导致mysqld崩溃了。终于承受不住。

我们来大概测算一下，100个并发需要多大的内存：

看来100并发线程，mysqld至少需要2G内存，另外考虑留给操作系统占用2G内存。所以一个4核8G机器，线程数不要设置超过250个。这个既是保护数据库不崩溃，保证响应时间在合理范围之内(1秒)，又是，当连接达到上限的时候，程序有报错，提示DBA需要增加机器的内存。

看完了这篇文章，相信你对"如何把mysqld压测到崩溃重启"有了一定的了解，如果想了解更多相关知识，欢迎关注行业资讯频道，感谢各位的阅读！