SQL Server 2012 : Performance Monitor Overview (part 4) - Working with Data Collector Sets

Working with Data Collector Sets

In addition to the provided Data Collector Sets, you can also create user-defined Data Collector Sets consisting of your own counters and settings. Real-time monitoring is great when a quick snapshot is required, but it can be difficult to identify patterns and trends when observing a server “live.” It’s usually more convenient to capture performance data to a file and then analyze that log file . This section walks through configuring a user-defined Data Collector Set to monitor system performance.

User-defined Data Collector Sets in Windows Server 2008 replace the Performance Logs and Alerts from Windows 2000/2003 but the principle is the same. To access them, from Reliability and Performance Monitor select Data Collector Sets. Right-click User Defined and choose New ⇒ Data Collector Set. A short wizard launches to create the new collector set. The first choice is to create from a template or create manually. Creating a collector set from a template provides three template collectors: Basic, System Diagnostics, and System Performance. You can use these templates as a starting point, adding and removing counters as required. Because these templates are Windows-generic, there’s nothing especially interesting about them from a SQL Server perspective. Therefore, choose the second option, Create manually (Advanced), and give the new collector a useful name, as shown in Figure 5.

FIGURE 5

The next step is to select Create Data Logs or Performance Counter Alerts. In most situations you’ll use the Performance Counter data log because you will likely be interested in gathering some system-wide performance data, rather than use PerfMon to fire an alert when a threshold is exceeded. Three types of data can be captured in the data log, as summarized in Table 2.

TABLE 2: Data Collector Set Logging Options

After selecting Create Data Logs, select the Performance counter log type and click Next to continue. Now you’ll add a small selection of interesting counters to get an overview of system performance. Click Add, and select all counters as shown in Figure 6. Leave the sample interval at 15 seconds; the impact of the sampling interval is covered in the next section.

FIGURE 6

After adding the counters, select a folder to store the trace data. Ensure there is sufficient space on the disk to hold the trace file (the size depends on how long the trace is running but normally 2GB of free space should be fine for a few hours of tracing). Click Next when you have entered a location for the logs. At the final confirmation dialog, click Finish to create the collector.

Starting, Stopping, and Scheduling Collectors

At this point, the data collector has been defined, but no data has actually been captured because the collector has not been started. To start a collector, right-click on the collector name and choose Start. Collectors with no stop condition configured will run until they are stopped manually. To stop the collector, right-click on the collector name and choose Stop. Collectors can be started and stopped as a whole, but performance logs or traces within a collector cannot be started independently of the collector. Define a new collector if this is required.

You can schedule collectors using the Schedule tab on the collector properties. When combined with a stop condition, both starting and stopping a collector can be fully scheduled.

Configuring Collector Properties

There are two points of interest in the properties dialog. One is the Directory tab, where you can change the folder used to store the log files. The other is the Stop Condition tab, which enables administrators to configure the duration of the collector — in seconds, minutes, hours, days, or weeks. Once the time configured in the stop condition has elapsed, the collector is automatically stopped.

Other points of interest on the collector properties dialog include the Schedule tab, which as it suggests enables administrators to schedule the start of the collector. There’s also a Task tab, where you can configure a task to run when the data collector stops, such as sending a MSG (new version of NET SEND) on completion.

Configuring Properties for Performance Counters

You may have noticed that there is no place in the collector properties to add or remove PerfMon counters — that’s because they are found in the Performance Counter properties. Because collectors can contain multiple data sources (listed in the right-hand pane), these properties are specific to each log type. Locate the Performance Counter log (usually named DataCollector01) and double-click it to show the properties.

Use the Performance Counter properties to modify log parameters, such as adding and removing counters, and changing log format and sample interval. The File tab contains further settings, including a checkbox to prefix log files with the computer name; this is particularly useful when comparing logs from multiple servers because it saves time opening files to identify the source server.

PerfMon Log Formats

There are four options for PerfMon log format: Comma Separated, Tab Separated, SQL, and Binary. The Binary log (BLG) type is the default and is suitable for most situations. Choosing SQL will require a data source name (DSN) to connect to SQL Server. There are some performance considerations when using this method because you want to limit the impact of monitoring to genuine users or server activity, and outputting trace data to the same instance being monitored is unlikely to help. Performance log files can be imported into a database post-capture for easier/better analysis, so avoid logging directly to SQL Server, unless there’s a good reason to do so and you’re confident you understand any impact on the monitored instance.

Using the Relog.exe tool it’s possible to manipulate log files, converting files between types; and if you’re working with large log files, you can narrow the time frame or extract some interesting counters.