What goes into a Maintenance Plan?
Unlike many other kinds of technology which generally stay static unless a change is intentionally applied, databases are always evolving and changing like a living being. And like all things that are alive, they need care and maintenance to stay productive and healthy.
Even for small databases, like those mandated by Express, database maintenance is a key part of continuous performance and the proactive reduction of system errors. A standard maintenance plan will address areas like:
- Index rebuilds / reorganizations
- Updating statistics
- Backing up the database
- Cleaning old backup files
- Purging unnecessary data
- System Data
- Application Data
Best Maintenance Plan Practices for SQL Server Express
For the most part, maintenance on database of 10GB or less is straightforward. Most systems are able to do index maintenance, update statistics, and purge old data extremely quickly on a data set of this size. The size restriction removes many potentially complicated scenarios where jobs must be run on individual partitions, by filegroups, or staggered across different days of the week.
Since SQL Server Express doesn’t have SQL Agent, and thus doesn’t have the GUI-focused maintenance plan feature at all, maintenance must be achieved using T-SQL. Luckily, most seasoned DBAs agree that T-SQL based maintenance plans are better than their GUI counterparts, so general SQL Server best practices are applicable to Express.
Ola’s Maintenance Scripts
Just as Ola Hallengrens free scripts are highly recommended for SQL Express backups, the same goes for his index and statistic scripts which help to make managing both areas extremely easy and customizable.
By default, Ola’s scripts use the following settings:
These translate to nothing being done to an index that is less than 5% fragmented; an attempt to reorganize, rebuild online, and rebuild offline (in that order) indexes fragmented more than 5% but less than 30%; an attempt to rebuild online and rebuild offline (in that order) if fragmentation is greater than 30%. Subsequent maintenance actions are only tried if the one before is unavailable or not supported on an index. For SQL Server Express, online index rebuilds are not available, so this option will always be skipped if not removed.
These numbers are based off of Microsoft’s BOL article Reorganize and Rebuild Indexes. Though that article may look new and shiny, most of its contents are from a many years ago and most aspects (like fragmentation thresholds) have never been updated. With that in mind, some higher thresholds than the default are not only acceptable, but encouraged.
A more modern take is setting the first fragmentation level to 30% and the second level to 80%. Some even prefer to go higher (see further reading at the bottom). Adjusting Ola’s defaults, by taking into consideration the lack of an online rebuild option and taking advantage of a few extra settings results in the following recommended script:
Since taking an index offline for a rebuild isn’t ideal, 80% makes sure that scenario isn’t occurring too often.
While normally a fill factor might need to be tweaked, it is probably safe to keep at 100% (no padding in the index) to make sure that the database files aren’t hitting the 10GB limit earlier than is absolutely necessary.
One of the great features of Ola’s script is that it will update statistics after it
does the index rebuilds, since an index maintenance operation will automatically
update statistics in use by the index. This prevents duplication of work and
unnecessary transactions. Using the
@OnlyModifiedStatistics flag further helps further by checking if a table’s data has been altered since the last time
statistics were updated. If the table hasn’t changed, then there is no reason to change
the statistics and it is skipped. This is especially helpful for archive tables or enum tables that
hardly or ever change, but may take up significant time to do statistics samplings on.
Logging to the built-in table is extremely useful when tracking down a failed index job or to analyze a growing trend in job runtime. It doesn’t cost much to log each action, but is key when hunting down problems with maintenance jobs.
Other Maintenance Tasks
Backup pruning is conveniently handled by settings within Ola’s script, so that part of maintenance can be bundled right into the backup commands themselves (see the Express Backup) article for details.
Data pruning is highly specific to an individual database’s needs, so usually needs to be developer a la carte. If you need help figuring this out for your databases, feel free to reach out via the contact page to get more one-on-one help.