Hi!
For efficiency the driver switch to vmalloc only after kmalloc fails. If the RAM on the server isn't suffcient this message will pop out frequently.
To avoid its apperance I suggest:
Tune the cgroup or docker settings to allow the driver use more memory;
Add more RAM to the server;
Switch to an older kernel, since this only happens after kernel 4.2;
Turn down the buffer size of the database. For MySQL/MariaDB using InnoDB it is the "innodb_buffer_pool_size" in the cnf file.
Now mysqld does not crash anymore, but we still see frequent page allocation errors. From syslog: