Comment 6 for bug 902654

Hi,
I am now able to reproduce this bug 100% of the time. I am now using 5.3.3 rc on linux.

I have attached a bundle which will allow you to do the same.

follow these instructions:

# create a db in mysql called "mydb"
# make sure that your ulimit -n is >= 8192 on your mysqld server due to the number of tables that are in this implementation
# unbundle mysql.bug.tgz
# use this script to populate the tables
$ cat aria_metric_data_tables0.sql | mysql -u <user> -p<pass> -D mydb
$ bin/mysqlreplay.sh <jdbcUrl> <user> <pass>
# e.g. bin/mysqlreplay.sh jdbc:mysql://localhost:3306/mydb myadmin mypasswd
# mariadb will crash

It turns out that the monitoring solution that we use calls

"show /*!50002 global */ status"
AND
SELECT * FROM information_schema.tables WHERE lower(table_schema) = 'mydb' AND engine is not null

by default this runs every 5 mins (same interval as the crash). When i change the interval on the monitoring solution to 20 mins, then mariadb crashes every 20 mins. When i turn it off the agent then mariadb is stable.

Please let me know if you are having issues with the script or are not able to reproduce. It works every time for me and the version of centos / rhel has nothing to do with it as i previously thought.

thanks.