My production system worked with mysql 4.1.x for about 2 years. For those 2 years we had about 3-5 table corruption problems. Hardware failure in most cases.
In January we switched to 5.0.28, and now we are getting around 1-2 corruptions per day.
I have 2 servers with Master-Save replication. Master gets ~200 qps and slave gets ~600 qps at the corruption moment. Top shows cpu load around 1.
In most cases corrupted tables happen with high delete/update/insert and only on Master server, Slave works perfect for all this month.
For testing purpose i wrote small script:
generate 15 table, with 4 field with about 50000 rec.
randomly send 11 type of query insert/update/select with 1,2,3 table
running this script on 3 different servers, shows that i get 1 crash in about 15 min 100%.
Trying different configuration, show that skip-concurrent-insert make test run for 3-4 hours without corruption, but when I leave it for a night, i get 1 table corrupted.
If I try different table creating options and different hardware configurations, i always get table corruption :(
Link to my script:
http://rus.leaping.net/mysql/test_db.pl
Link to my.cnf:
http://rus.leaping.net/mysql/my.cnf
Every suggestion is appreciated!
Thanks!