Modifying the number of similar requests

This topic explains how the system can be configured to reduce the number of similar requests it collects.

All Server Request instances are counted and measured for latency violations and errors, and represented by general metrics. However, only certain Server Request instances are monitored along with the Call Tree for extended viewing.

For a Server Request to be monitored it must first pass the sampling mechanism, and then have a latency violation, or have detectable reference in the HTTP header (meaning that another BMC product has requested monitoring), or have an Exception or Code Fault.

The count to latency violation and error rate must be closely watched and should represent a Service Level percentage for the product.

The system uses the following parameters in the Agent policy file to define the number of similar requests it collects:

Where

• <sla_breach_time> is the time, in milliseconds, in which only the worst SLA breaching requests will be persisted.
• The default is 300000 milliseconds (5 minutes).

Where

• <timeframe> is the time, in milliseconds, in which error requests are buffered.
• The default is 300000 milliseconds (5 minutes).

Where

• <number> is the number of different error types that will be sent in the timeframe frame specified in the persisting.error.server.requests.timeframe parameter.
• The default number of errors is 3.

Where

• <threshold> is a global static threshold, in milliseconds, used for all entry points.
• The default is 2000 (2 seconds).

Changing these parameters results in the BMC Application Diagnostics Agent storing more incidents of the same erroneous request.

Take note that this configuration restricts the number of similar errors as follows:

• No more than the number specified in the persisting.max.num.unique.errors.per.timeframe of different error requests will be sent to the Collector in the specified time frame.
• A specific error request will not be sent more than once to the Collector in the specified time frame.
• An error request will be sent to the Collector only if it occurred in the previous time frame.
• If more than persisting.max.num.unique.errors.per.timeframe error types occur concurrently for a while, they will be sent to the Collector in a round-robin fashion, giving representation to all error types, without violating other restrictions in these parameters.