I'm using mysql 4.1.14 rpm's on Fedora Core 4. I've setup max_heap_table_size to allow for 500M heap tables. Currently I use roughly 435M in heap by about 16 different tables. Some use btree, other's use default hash.
The problem is, over time mysql looks like this in top: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 2898 mysql 15 0 1127m 626m 2076 S 18.3 62.5 2974:17 mysqld-max mysql is forcing the machine to swap. I've tried dropping all heap tables that are in use and this does not free the ram. I have to restart mysql and and using init-file reload the heap tables and everything is fine for a couple of days. My problem is strikingly similar to this: http://groups.google.com/group/mailing.database.myodbc/browse_thread/thread/485647dae02b59a1/32f0009e9cb135b7?lnk=st&q=virtual+heap+mysql&rnum=1&utoken=rT79JjoAAABobq0US6-f3p1tupn-bp7-GyqMAsXdt4_lvPhOluyGzfrEz8xuJ8FzZhQCB5gw1_s38laLLlcPg_ShAKo-q_vP I'm using the mysql rpm's from mysql.com <http://mysql.com>. Other than restarting mysql nightly, is there anything I could do to further debug this? Thanks, Mysql Rocks.