You can only view Server Visibility metrics in the Server Visibility version of the Metric Browser. You see this when you access the Metric Browser from the Servers tab in the top navigation bar of the Controller UI.

Machine Agent Versus .NET Machine Agent

If a server has both Machine and .NET Agents installed, there may be differences in metric values reported by the Machine Agent and the .NET Agent due to different averaging rates and measurement methods. See .NET Compatibility Mode.

Streamlined Browsing Mode

The Server Metric Browser includes a Streamlined Browsing option for node metrics. Use this option to browse metrics for tiers that contain multiple nodes.

• With Streamlined Browsing enabled, each node in the browser tree view shows only metrics reported by the Machine Agent on the corresponding node.
• With Streamlined Browsing enabled, the Hardware Resources and Custom Metrics folders do not display.
• With Streamlined Browsing disabled, each node in the browser tree view also shows metrics for other nodes in the same tier (with no indication of the node that corresponds to which tier group).

This streamlining option is not enabled by default. Splunk AppDynamics recommends that you enable this option. To enable this option, go to:

http://<controller host>:<port>/controller/admin.jsp and set sim.metrics.metricBrowser.machineMetricMappings.enabled to true.

Linux Kernel Processes

The Machine Agent has a regex processSelectorRegex setting that specifies the set of processes monitored by the agent. The default

Default Monitoring Mode

TheDefault Monitoring Modecolumn indicates the default category of each metric when Dynamic Monitoring Mode (DMM) is enabled. When this mode is enabled, a Machine Agent reports metrics based on the DMM setting on that server:

• KPI – Report Key Performance Indicator metrics only
• Diagnostic – Report KPI and Diagnostic metrics
• Advanced – Report all unfiltered metrics on the Machine Agent.

Load Average Percentages are Scaled by 100 in the Metric Browser

Load Average metrics are shown as float values (such as 0.70 or 1.05) in the Server Dashboard. In the Metric Browser, these metrics are multiplied by 100 to provide two decimal points of precision. If a server has an average load of 7.67, for example, the Server Dashboard shows the Load Average as 7.67 and the Metric Browser shows the Load as 767. To use one of the metrics in a health rule or custom dashboard, divide the metric by 100.

Stolen Times are Reported Differently

On Windows Machine Agents, if CPU %Stolen is not matching values reported by Splunk AppDynamics, it may be because Windows Performance Monitor's (Perfmon) counters operate at a granularity of 100ns for CPU metrics. For CPU %Stolen, the counters are provided by the Hypervisor, which operates on a granularity of 1ms. Windows Perfmon divides the CPU %Stolen counter (operating on 1ms) directly by time (in units of 100ns), which results in the CPU %Stolen values being reported at a very low percentage (~0.01%).

Splunk AppDynamics makes the correction between different units of time, which is why values reported by Splunk AppDynamics are greater than Perfmon's values by a factor of 10,000 (time in 1ms * 10,000 = time in 100ns). CPU %Stolen values reported by Splunk AppDynamics sometimes exceed 100% under high load. These CPU %Stolen values are a result of multiple cores being used by the guest machine, where %Stolen time is added across multiple cores.

To compute the %Stolen time:

1. On a PowerShell terminal in the monitored guest machine, run Get-WmiObject Win32_PerfRawData_vmGuestLib_VCPU.
2. Note the CpuStolenMs counter (in ms).
3. Note the Timestamp_PerfTime counter (in ms, take a look at Frequency_Perftime, this value should be 1000 (in hz)).
4. After a minute, run Get-WmiObject Win32_PerfRawData_vmGuestLib_VCPU again and take note of both the counters once more.
5. CPU %Stolen = 100% * (CpuStolenMS (at T2) - CpuStolenMs (at T1) / (Timestamp_PerfTime (T2) - Timestamp_PerfTime (T1)).

HardwareMonitor and JavaHardwareMonitor Calculate "Writes/Sec" Differently on AIX Machines

The Machine Agent calculates the Disk metric writes/sec on AIX machines differently, depending on whether the HardwareMonitor or the JavaHardwareMonitor extension is used. These extensions assume different block sizes when calculating these metrics. For HardwareMonitor, the block size is 4096 bytes. For JavaHardwareMonitor (based on SIGAR), the block size is 512 bytes. Because the number of writes/sec is calculated as <kb_written>/<block_size>, the HardwareMonitor results are lower than those calculated by JavaHardwareMonitor.