get a lever and pick any turtle: lifting with metadata

G E T A L E V E R A N D P I C K A N Y T U R T L E

L I F T I N G W I T H M E TA D ATA

EMA I L

WEB

Get a Lever and Pick Any Turtle – Lifting with Metadata Rose Crescent | 3

Give me the place to stand,

and I shall move the earth.


Start Simple


Scale

the

Technique


Use

Basic

Concepts


Expand

to

System

s


Mixing Metaphors –

We’re Going Down a

Rabbit Hole

• Pam: And then outta

that cake pops another

stripper holding a

smaller cake and then

an even smaller

stripper pops outta that

one.

• Michael: What is that

smaller stripper

holding?

• Pam: A cupcake! It's

cupcakes and

strippers all the way

down.


Avoid Injury


Metadata has always been there

Codd Rule #0: The RDBMS

must use the relational

facilities to manage the

database


Metadata has always been there

Codd Rule #4: There must be an

active online catalog based on the

relational model


DBMeta

Demo

DBHealt

h


Tables

SPs

Views


• INFORMATION_SCHEMA contains tables and

views (ANSI 92, kind of generic RDBMS model,

doesn‟t expose a lot of implementation details)

• sys schema contains tables and views (more

proprietary, particularly non-portable things like

indexing features)

SQL Server Basic Metadata Services


• You could make tables of additional information about database objects

• DON’T

• Built-in extended properties work great for this!

• MS_Description: Microsoft‟s standard “description” property

• sp_addextendedproperty, sp_updateextendedproperty, sp_dropextendedproperty

• fn_listextendedproperty

• sys.extended_properties

• I‟ll improve this in the DBMeta schema

SQL Server Extended Properties


Layer this view on top of sys.extended_properties

to:

• Make it easier to use

• Give a standard naming convention

• Easier than CROSS APPLY

fn_listextendedproperty

DBMeta.Properties


• Easier to use than sp_addextendedproperty,

sp_updateextendedproperty, sp_dropextendedproperty

• The builtin procs have some awkward hierarchy aspects:

e.g. a column or trigger is defined at level 2 where the table

is level 1

DBMeta.AddXP, UpdateXP, DropXP


• With some simple tools we now have an easier way of dealing with the metadata in the extended properties

• This can be expanded at will to join regular base metadata information, PIVOT with the property information and combine them in interesting ways

• For instance, say your question was “I want to see NULLability (base metadata) of all money (base metadata) columns in tables in the Accounts subsystem (extended property) which have not been marked as reviewed. (extended property)” You can do that.

DBMeta


• Some questions about the health of the database system will be very standard

• By health, I mean outside of integrity you can realistically enforce with constraints and beyond what you might use DDL triggers for

• Can we use the metadata to do something generally useful?

• Can we organize the health system using metadata so that it is self-maintaining as much a possible?

DBHealth


• Let‟s create some data structures in

our Demo schema – some tables

and indexes

Demo Data


• Arbitrary, but useful and chosen for variety

• Rule #1: All tables should be identified with a SUBSYSTEM (Organizational Rule)

• Rule #2: No unique indexes with columns which allow NULLs (Indexing Rule)

• Rule #3: No varchar(N) columns where N <= 2 (Table Design Rule)

• Rule #0: We only want to enforce these on schemas under management

Let’s Make Up Some DBA Rules


• We simply make a view DBHealth.MonitoredSchemas

• We‟ll need to remember to use this view

• You could make a whole layer of views which filter through this, of course

Rule #0


• Look for tables without extended

property „SUBSYSTEM‟

Rule #1


• Look for unique indexes where any

of the columns in the index are

NULLable.

Rule #2


• Find all varchar/nvarchar columns with

length <= 2

Rule #3


• Let‟s tag things for exclusion with

extended properties!

What about exclusions?


• How do we organize them?

• Could we put them in individual views or procedures to make them easy to use?

• Then make a master procedure which runs them all.

• But we‟d need to make a list of all the procedures

• But what did we say?

• Let‟s use the metadata to tag the procedures!

• We‟ll mark the procedures

• We‟ll categorize the procedures

• We‟ll be able to look for all the marked procedures and run them in an automated fashion – no new tables, limited maintenance!

Now we have a bunch of rules


• All procs in DBHealth should have

HEALTH_CHECK_PROC and

HEALTH_CHECK_SET properties

• Show me any unindexed tables (heap tables with

no non-clustered indexes)

• Show me any items tagged with a „TODO‟

extended property

• Show me all columns in the database which

aren‟t actually used (yes, you can do this!)

Ideas For Rules


• Health Monitoring

– .NET

• Code Generation

– T-SQL

– T4

Let’s use this from a client


Materials

• http://bitly.com/bundles/caderoux/3

– 4 Part article by Adam Aspin on SQL Server Central

– Article by Brent Shaub on MSSQLTips.com

– Google Code Repository of presentation and code/scripts

http://bitly.com/bundles/caderoux/3

http://bitly.com/bundles/caderoux/3

EMA I L

WEB

get a lever and pick any turtle: lifting with metadata

Technology