MariaDB 10.3.30 to 10.6
Hi Support,
Recently we done MariaDB 10.3 upgrades to 10.6. We done before and after checks on SQLs like CREATE INDEX, UPDATE TABLES, SELECT COUNT(), etc. However, those sqls which took not more than 1mins for table size 74gb (with index size) is not even starting updates in 1hr. We tried everything that mentioned in similar issue ticket raised by other members. 1) ANALYSE TABLE 2) OPTIMISE TABLE 3) change conf value such as optimizer_search_depth=0.
Our server.cnf file has been modified as per previous required configuration.
Kindly let us know how we can resolve it.
[mysqld] sql_mode=NO_ENGINE_SUBSTITUTION #skip-grant-tables #innodb_force_recovery = 1 port = 3306 socket = /tmp/mysql-prod.sock log-error = /var/log/mysql/error-prod.log datadir = /mysql/data basedir = /usr query_cache_type = 2 skip-name-resolve skip-external-locking ft_stopword_file = "" ft_min_word_len = 1 table_open_cache = 10k table_definition_cache = 10k max_heap_table_size = 512M tmp_table_size = 512M open_files_limit = 80000 join_buffer_size = 50M sort_buffer_size = 4M read_buffer_size = 4M read_rnd_buffer_size = 16M max_allowed_packet = 512M max_connections = 1500 thread_stack = 500k thread_cache_size = 8k query_cache_limit = 200M query_cache_size = 256M concurrent_insert = ALWAYS bulk_insert_buffer_size = 25M group_concat_max_len = 1000000 slow-query-log-file = /backup/logs/prod-replica-slow.log slow_query_log = 0 log_queries_not_using_indexes = 0 # Added Newly long_query_time = 0.15 # reduced from previous value 0.25 local-infile = 1 low_priority_updates = 1 wait_timeout = 7200 thread_concurrency = 48 character-set-server=utf8 collation-server=utf8_general_ci net_read_timeout=90 net_write_timeout=90 interactive_timeout=300 connect_timeout=90 plugin-load-add = file_key_management file_key_management_filename = /mysql/test/fullkeyfile.key file_key_management_encryption_algorithm = AES_CTR innodb_data_home_dir = /mysql/data innodb_data_file_path = ibdata1:10M:autoextend innodb_log_group_home_dir = /mysql/data innodb_buffer_pool_size = 10G #50G innodb_flush_log_at_trx_commit = 2 innodb_file_per_table = 1 innodb_flush_method=nosync #O_DIRECT innodb_log_file_size = 48M innodb_log_buffer_size = 16M #skip-federated
Answer Answered by Daniel Black in this comment.
Show the SQLs, the SHOW CREATE TABLE, and EXPLAIN query.
Enable your slow query log. Consider explain-in-the-slow-query-log. log_queries_not_using_indexes will be noisy if min_examined_row_limit isn't set to 1000 or so to reduce noise.
Recommend stripping your config back to the items you've explicitly set because of outcomes particularly disabling query cache.