One time MySQL paralyzed rescue

The recent increase in project access traffic, especially after a new business message push, the Aurora pushed to the phone to receive 190,000 clients, the app immediately began to access the speed is slow, the user experience is quite poor

Customer service received the question of the phone start one by one, I look at the end of the thought, it must be the flow up, or the database is either nginx, or bandwidth is not enough, only these three kinds of possible

Solve the problem quickly

Open Database Service monitoring: The number of process connections reaches 1200, and each process has SQL in process. Such as:

Obviously the problem comes out,MySQL database is not up, I use the MySQL branch Percona, anti-high concurrency is stronger than the official version Https://www.percona.com/docs/wiki/benchmark:mysql : 554-tpcc:start

Then adjust the parameters, then restart, the middle of the restart occurred an error, I was frightened, and then hurriedly restore my. CNF, fortunately every time the change will be back up, or only cry

Restore the configuration, MySQL can start, but has been started immediately 1200 connections are all stuffed, the app is still slow, my God, do not play with me,

Then adjust the parameters and then do not re-enable the service mysqld reload command to operate

[mysqld]basedir=/usr/local/mysqluser=mysqlsocket=/var/run/mysqld/mysqld.sockserver_id=1local_infile=1tmpdir=/ Mnt/fiodatadir=/mnt/fio320skip-grant-tableinnodb_buffer_pool_size=4g[Real memory 8G]innodb_data_file_path= Ibdata1:10m:autoextendinnodb_file_per_table=1innodb_flush_log_at_trx_commit=1innodb_log_buffer_size=16minnodb_ log_files_in_group=2innodb_log_file_size=900minnodb_thread_concurrency=0innodb_flush_method             = O_ DIRECTinnodb_write_io_threads=9innodb_read_io_threads=9Innodb_io_capacity=500innodb_max_dirty_pages_ Pct=90max_connections=12000query_cache_size=0skip-name-resolvetable_cache=400

Adjusted after the reload, the instant queued SQL slowly decreased, the process began to gradually reduce, to 130 stable
Open the app again, the same speed as usual, with mysqlworkbench monitoring, hit rate and buffer usage reached 100%

One time MySQL paralyzed rescue