Best Practices Every DBA Must Know
By Brad M. McGehee
Installing & Upgrading SQL Server
* Generally, when installing a new SQL Server instance:
• Use the newest hardware firmware and OS drivers.
• Use the newest OS version with latest SP and patches.
• Use the newest SQL Server version with latest SP & Hot Fixes.
• Test, and once stable, put into production & be wary of making changes.
* Generally, when upgrading an existing SQL Server instance:
• Don’t upgrade unless you have a good reason to upgrade. If your instance is working well, don’t mess with it.
• For example, upgrade if you need new features, or have problems with an old installation, or need to upgrade hardware.
• It is always safer to upgrade to a new server with a fresh installation of the OS and SQL Server than to upgrade in place. This allows you to test more effectively, and also gives you an easy “back out” option.
Security Basics
• Don’t give users more permissions than they need to perform their job. (Critical. Sounds simple, often hard.)
• Don’t use the SA account for anything. Assign it a complex password, and keep it handy just in case. Instead, use a domain account that is a member of the sysadmin role.
• Don’t allow an application to use the SA or a sysadmin account to access SQL Server.
• Use Windows Authentication security whenever possible. (Applicable for in-house development).
• Don’t give vendors sysadmin access to your servers.
• Log off or lock your SQL Server (or workstation) when done.
General Server Configuration
• Ideally, SQL Server instances should run on a stand-alone server (physical or virtual) with no other major apps running on it. Small monitoring apps or utilities are generally OK.
• Avoid multiple instances unless you have a really good reason to use them. Consider virtualization instead.
• Unnecessary SQL Server services should be uninstalled or turned off.
• Ideally, don’t run antivirus/antispyware software locally. If your organization’s policy requires running antivirus/antispyware software locally, exclude MDF, NDF, LDF, BAK, TRN and BCP files.
SQL Server Property Settings
• SQL Server includes many instance-wide property settings. (e.g. sp_configure)
• Don’t change any of these defaults unless you know exactly what the effect of the change is.
• There are a few of these settings that are commonly changed from their defaults, and generally you need to perform
baselines before and after the change to see if performance and HA increased or decreased because of the change.
• If you make a change, only make one change at a time so that you can see its effect independent of other variables.
Memory Configuration
• Ideally, use the 64-bit version of the OS and SQL Server.
• If using the 32-bit version of SQL Server, and if using 4 GB or more of RAM, ensure that /3GB switch and AWE memory are correctly configured. Correct settings depend on available RAM.
Data and Log File Management
• Remove physical file fragmentation before creating new or expanding existing MDF or LDF files.
• When creating new MDFs and LDFs, pre-size them to eliminate/minimize autogrowth events.
• MDF files should be located on their own disks.
• LDF files should be located on their own disks.
• BAK and TRN backup files should be located on their own disks.
Instant File Initialization
• Enable instant file initialization, which prevents MDF files from being zeroed out when they are grown, allows MDF files to be created quickly. LDF files are not affected.
• Speeds up CREATE DATABASE, ALTER DATABASE, RESTORE DATABASE, Autogrowth.
• Requires SQL Server 2005/2008, and Windows Server 2003/2008 (or higher version).
• Instant file initialization is turned on if the SQL Server (MSSQLSERVER) service account has been granted the SE_MANAGE_VOLUME_NAME permission by adding the service account to the Perform Volume Maintenance Tasks security policy. Members of the local Windows Administrator group automatically have this right.
Don’t Shrink Files
• If you properly size your MDFs and LDFs, then you should very rarely have to shrink a file.
• Don’t schedule periodic database or file shrinking operations.
• If you must shrink a database (MDF file):
– Do so manually
– Rebuild the indexes after the shrink is complete
– Schedule these steps during the slow time of the day
• Benefits of not automatically shrinking files:
– Eliminates grow and shrink syndrome
– Reduces physical file fragmentation
– Reduces resources used for these operations, allowing more important tasks to use them
Tempdb Management
• Pre-size tempdb so autogrowth doesn’t have to happen often (8MB is default, which is very low).
• Set autogrowth to avoid many growth spurts, use a fixed amount that minimizes autogrowth use. (10% is default,
which causes lots of autogrowth).
• If tempdb is very active, locate it on its own disks.
• If very active, consider dividing the tempdb into multiple physical files so that the number of files is 1/4 to 1/2 the number of CPU cores, up to 8 files. Each physical file must be the same size.
Database Property Settings
* Don’t change database property settings unless you have a very good reason. Some key ones:
• Auto Create Statistics: On
• Auto Update Statistics: On
• Auto Shrink: Off
• Autogrowth: Leave on. Use mainly for catching mistakes. File growth should be managed manually. Use fixed amount that minimizes autogrowth occurrences.
• Recovery Mode: Set to full for all production databases so transaction log backups can be made.
• Page Verify: Use Checksum (2005/2008), don’t turn off.
• Compatibility Level: Should be set to match current server version, unless there are compatibility problems.
Configuring Jobs—General
• If your production servers don’t have any jobs, then there is a problem, as all servers need jobs.
• Try to schedule jobs so they don’t interfere with production.
• Try to prevent jobs from overlapping.
• Set alerts on jobs so you are notified if they fail.
• Check jobs daily to verify that they have run correctly (not hung, not run abnormally long, etc).
• If you use the Maintenance Plan Wizard, be careful to use it properly. If misused, it can create maintenance jobs that hurt performance.
Create Index Rebuilding/Reorganize Job
• Indexes need to be rebuilt or reorganized regularly to minimize fragmentation and reduce wasted space.
• Only Reorganize or Rebuild indexes that need it, don’t defrag all indexes all the time (unless perhaps a database is small).
• Some general recommendations (from BOL). Use as a starting point to determine what is best for your server.
– If fragmentation is less <5 alone.="alone." leave="leave" p="p" then="then">– If fragmentation is >5% and <30 consider="consider" p="p" reorganize.="reorganize.">– If fragmentation >30%, consider Rebuild.
• Use sys.dm_db_index_physical_stats to help you determine if an index should be rebuilt or not.
• Reorganize or Rebuild jobs should ideally be scheduled as a SQL Server Agent job using a custom T-SQL script you create to meet your environment’s specific needs.
• Perform as needed to minimize negative effect of index fragmentation, but not more than required.
Create Data Corruption Detection Job
• Ideally, run DBCC CHECKDB as frequently as practical.
• If you have a problem, you want to find it as soon as possible to reduce the risk of data loss.
• Create an appropriate job to run this (or similar) command:
DBCC CHECKDB ('DATABASE_NAME') WITH NO_INFOMSGS, ALL_ERRORMSGS;
Note: Consider using PHYSICAL_ONLY option for large or busy production servers to reduce run time.
• Don’t use the DBCC CHECKDB repair option unless you fully understand its implications, as data loss will most likely occur.
Implement a Backup/Restore Strategy
• Create a job to perform full backups daily on all system and user production databases, plus log backups hourly (or similar variation). Include differential backups if appropriate.
• If a database uses the bulk or full recovery model, you must back up the transaction log to keep it from growing uncontrollably.
• Backup using RESTORE WITH VERIFYONLY to help verify backup integrity. (Does not guarantee good backups.)
• Periodically test backups to see if they can be restored.
• Set up an appropriate backup retention policy.
• Store backups securely and off-site (not on same disk array or SAN).
• If you have a limited backup window, or have limited disk space, use backup compression. Can be a big time saver.
Set Up Alerts for Critical Errors
• Create a SQL Server Event Alert for all events with a severity of 19 [fatal] and higher.
• Have alerts sent to you or whoever is responsible for day-to-day monitoring.
• Consider a third-party alerting tool if SQL Server Alerts doesn’t meet all of your needs.
Create a Disaster Recovery Plan
• You must create a document that outlines, step-by-step, in detail, how you will recover your SQL Servers in the case of any problem, small or large.
• You need to practice using the plan so you are familiar with it and can easily implement it.
• Keep Microsoft SQL Server’s Product Support phone number handy. Paste it near your computer.
• Remember: Most “disasters” are small, such as a corrupted database. Big “disasters” occur very rarely, if ever. But you need to be prepared for both.
Document Everything
* Yes, documentation is very boring, but it is very critical to being a successful DBA. Be sure to document:
• The installation and configuration of each instance.
• The installation and configuration of any application that uses SQL Server as its back end (as related to SQL Server).
• Troubleshooting tasks, as the same problem may reoccur, and you don’t want to reinvent the wheel.
• Any time any change is made to any instance for any reason.
* Be sure that documentation is easily available to everyone who needs access to it.
Test Everything
• Before you make any change on a production SQL Server, be sure you test it first in a test environment.
