character-set-server = utf8mb4 #tunning skip_name_resolve = 1 open_files_limit = 65535 max_connect_errors = 1000000 table_open_cache = 1024 table_definition_cache = 1024 #table_open_cache_instances = 64 #MySQL 5.6 sort_buffer_size = 4M join_buffer_size = 4M interactive_timeout = 600 wait_timeout = 600 tmp_table_size = 32M max_heap_table_size = 32M #master_info_repository = TABLE #MySQL 5.6 #relay_log_info_repository = TABLE #MySQL 5.6 #gtid_mode = on #enforce_gtid_consistency = 1 read_buffer_size = 8M read_rnd_buffer_size = 4M bulk_insert_buffer_size = 64M lock_wait_timeout = 3600 #explicit_defaults_for_timestamp = 1 #MySQL 5.6 ...

http://www.uuread.cc/cfaqxcQc.html Kubernetes具備超強的故障恢復機制，Kubernetes會重啟意外停止的pod。我們曾遇到過因記憶體洩露，導致容器在一天內宕機多次的情況，然而令人驚訝的是，甚至我們自己都沒有察覺到。 資料儲存 以我們的情況，在Kubernetes內執行資料儲存沒有想象中那麼完美，設定起來也比其他Kubernetes部署複雜得多。 於是我們決定不在生產環境資料儲存上使用Kubernetes，而是選擇在不同的機器上手動啟動這些叢集，我們在Kubernetes內部執行的應用正常連線到資料儲存叢集

http://tech.noredink.com/post/133347435578/optimizing-mysql-for-high-concurrency-on-amazon Looking for concurrency bottlenecks performance schema List the total maximum wait by event inside of InnoDB SELECT EVENT_NAME, SUM_TIMER_WAIT/1000000000 WAIT_MS, COUNT_STAR FROM performance_schema.events_waits_summary_global_by_event_name ORDER BY SUM_TIMER_WAIT DESC, COUNT_STAR DESC LIMIT 30; List current and last wait events on InnoDB. -- “what the current, or last completed, wait for each session was, and for exactly how long they waited” SELECT NAME, IF(PPS.THREAD_ID = 0, ‘Internal Thread’, CONCAT(IPS.USER, ‘@’, IPS.HOST)) USER, DB, COMMAND, STATE, TIME, EVENT_NAME LAST_WAIT, IF(TIMER_WAIT IS NULL , ‘Still Waiting’, TIMER_WAIT/1000000000) LAST_WAIT_MS FROM performance_schema.events_waits_current JOIN performance_schema.threads PPS USING (THREAD_ID) LEFT JOIN INFORMATION_SCHEMA.PROCESSLIST IPS ON IPS.ID = PPS.THREAD_ID; ...

https://medium.com/devoops-and-universe/database-performance-aws-vs-bare-metal-452f64481b07

https://mariadb.com/kb/en/mariadb/performance-schema-overview/