historical: http://lists.mysql.com/mysql/226361 after it turned out to be stable in 2013/2014 upgrade to MariaDB 10.0.17 yesterday and the logs started with "table corrupt", "table not closed properly" and so on for tables which where not only read this time not with 'mysqld got signal 11' but "only" corrupted MyISAM files and after disable 'myisam_use_mmap' all is clean and stable over hours BTW a 'repair table' don't delete the TMM file leading in a 'optimize table' later failing until you delete the file manually from disk __________________________________________________________________ this configuration is unchanged for years and so my only conclusion is that Oracle managed to kill the feature again, this is also confirmed for Percona Server 5.6 where even queries fail confirmed from another sysadmin which found out that playing with my settings leaded to problems while here MariaDB 5.5.x was in use and hence no troubles - given that 'myisam_use_mmap' is borked at MariaDB 10.x and Percona 5.6 it's likely a backport from MySQL 5.6 delay-key-write = ALL concurrent_insert = 2 open-files-limit = 750000 myisam-recover-options = "FORCE" myisam_use_mmap = 0 wait_timeout = 300 interactive_timeout = 300 max_allowed_packet = 200M max_connections = 500 max_tmp_tables = 150 max_connect_errors = 250 max_delayed_threads = 32 flush_time = 0 query_cache_limit = 256K query_cache_min_res_unit = 1K query_cache_size = 1280M query_cache_type = 1 table_cache = 10000 table_definition_cache = 10000 thread_cache_size = 400 tmp_table_size = 256M max_heap_table_size = 256M key_buffer_size = 256M sort_buffer_size = 256K myisam_sort_buffer_size = 15M join_buffer_size = 1M preload_buffer_size = 128K read_buffer_size = 128K read_rnd_buffer_size = 128K