Top 10 OPERATIONAL Mistakes TO AVOID - reservoirsw.comreservoirsw.com/wp-content/uploads/2012/03/Top-10... · Top 10 OPERATIONAL Mistakes TO AVOID Helping keep your Sanity Intact

Top 10 OPERATIONAL Mistakes TO AVOID

Helping keep your Sanity Intact

Mike Hillwig

�  AKA The Cranky DBA

�  SQL Server DBA

�  Working with SQL Server since SQL 7

�  Specializes in Operational Issues and Scalability

�  Contract DBA at hosting division of a financial software company

�  Resume includes Acme Packet, Shawmut Design and Construction, Equitable Resources

Mike Hillwig

�  crankydba.com

�  sqlserverpedia.com

�  twitter.com/mikehillwig

Avoiding Messes

Before We Begin A little common sense advice…

Beware of Advice from Random Blogs

�  Anybody can put bad advice on the internet

�  Trust People You Know

�  I Don’t Trust Anyone Who Says “ALWAYS” or “NEVER”

�  If it comes from Paul or Kimberly, it’s true.

Top Ten Rules

#10: Not Planning to Scale or Grow

�  “It’s just temporary”

�  “We only need it for proof of concept.”

�  “Can’t we use the existing server?”

�  “We’re just going to demo this product.”

“Oh, by the way…”

�  “Our sales volume just doubled.”

�  “Didn’t we tell you?”

�  “We just bought a competitor.”

�  “We’re now running the business with that product.”

�  “You’re backing that up, right?”

#9: Not using a Domain Account for SQL Services

�  Allows you to interact with other servers. Very helpful with backups over the network.

#8: Not Configuring SQLMail Alerts

�  Allows you to know when something fails

�  Allows you to be active instead of reactive

�  Not taking advantage of free monitoring

#7: Not Allowing Instant File Initialization

�  Prevents rapid growth and creation of database (or any) files

#6: Using your Server as a Workstation

�  Avoid Remote Desktop

�  Never use SSMS directly from your server

�  Instead: Run from a workstation

�  RUNAS is your friend

#5: Improper Use of Traces

�  Avoid in Production

�  Creates more work for your server

�  Never store on your Database Server

�  Instead: Run from a workstation and store on another machine

#4: Allowing Data Files to Grow in 1 MB or 10 Percent Increments

�  Defy the Defaults!

�  Causes disk fragmentation

�  Every growth will ALTER the database

�  Disk is cheap!

�  Causes high number of VLFs in transaction log files

�  Instead: Grow in larger increments

#3: Not Testing Restores

�  The worst time to test a restore is when your production server fails

�  It’s good practice for when something does fail

�  Don’t forget to test recovering from long-term storage

#2: Not Setting Min and Max Memory

�  MIN: ”Use this much so nobody else can have it.”

�  MAX: “Save some for the Operating System” – Prevents Paging

�  Very important in virtual environments

#1: Enabling Auto Shrink

�  You can’t control when it runs

�  What you shrink will just grow again

�  Fragments your indexes

�  Instead: Use DBCC Shrinkfile

Bonus Round

Not Configuring Model Database

�  This is a template for newly created databases

�  Helps enforce standards

Keeping System Databases on OS Drives

�  Major performance killer

�  Avoid putting TempDB and the Windows Pagefile on the same spindles/LUNs (Notice I didn’t say “never”)

�  Avoid using default file paths in the SQL Installer

Not managing virtual log files inside transaction log files

Schema Changes at Peak Times

�  ALTERing TABLEs causes locks on the entire table

�  High Risk

�  Difficult to roll back at peak times

�  Instead: Use a Maintenance Window

For More…

�  crankydba.com/go/topten

�  twitter.com/mikehillwig