There are many brain-dead things that I've seen in databases over
the years, worst practices that are guaranteed to cause data loss and
unplanned outages.
Also see
Oracle worst
practicesThere are some of the major areas for database
worst practices, generic to all databases.
Suboptimal Database Design
Back in the 1980's when disk was $200k/gig, designing
a database in third normal form was a perfect approach, since 3NF is
non-redundant, and a DBA management goal was to save expensive disk
space.
The rules of normalization are required to understand the
relationships & functional dependencies, but BCNF (or 3NF for those
non-purists) is just a starting point, not a completed model and it's a
database worst practice to design tables without reduncandy to reduce
runtime table joins..
Inadequate Indexing
One of the top causes
of excessive I/O during SQL execution is missing indexes, especially
function-based indexes, and failure to tune the instance according to
the SQL load is a major worst practice.
Poor Metadata
Management
The
database worst practice was to re-analyze the schema metadata on a
schedule, forgetting that the purpose of re-analyzing schema statistics
is to change your production execution plans. This worst practice
has become so commonplace that it has been dubbed the "Monday Morning
Surprise".
Shops with strict
production change control procedures forget that analyzing the
production schema can effect the execution of thousands of SQL
statements.
