Foglight 7.3.0 - Performance Tuning Field Guide

Viewing the Settings that Control the Size of Stored Data

The default time window during which the host collects and stores data that will be accepted by the server after the connection is restored is one hour. This setting can be changed to a desired value using a VM option in a configuration file on the server side. The Foglight® Agent Manager keeps track of this time window and discards any messages that are collected and stored on the host outside of this time window. For more information about changing the length of the time window for store and forward processing, see Store and Forward Processing and the Foglight Management Server .

The size of the stored data is controlled with a set of parameters in the file <foglight_agent_manager_home>/state/default/config/fglam.config.xml. In that file, the <config:queue-sizes> element controls the amount of the stored data using the following attributes of its <config:upstream> and <config:downstream> sub-elements:

For example, the default configuration instructs the Foglight Agent Manager to reserve 1 MB of disk space for storing messages and to limit the number of messages in a batch to 500, for upstream connections:

NOTE: The max-queue-size attribute is used internally and should not be changed.

CAUTION: The above attributes illustrate a default configuration. Changing them is only recommended when instructed to do so by Quest Support.

IMPORTANT: Changing the default settings of the above attributes is only recommended when instructed to do so by Quest Support.

Examples

This contains scenarios, each illustrating a set of common results typically found in store and forward processing:

Example 1: Reporting on Disk Usage

A Foglight® Management Server monitors a group of hosts, each using the Foglight Agent Manager to communicate with the server. The network between a monitored host and the server becomes unavailable at 4:55 PM, and is restored at 5:00 PM. Prior to the connection disruption, the monitored host’s disk usage was within a normal range.

At 4:55 PM, the connection with the server is lost, and the Agent Manager process continues to collect the data from the host, and to store it on disk.

At 4:57 PM, a high disk usage is reported. The Agent Manager stores this information on disk.

At 5:00 PM, the network connection between the Management Server and the Agent Manager is restored. The Agent Manager forwards the data stored between 4:55 and 5:00 PM to the Management Server. The increase of disk usage is reported at 5:00 PM, when the data arrived, not at 4:57, when it actually occurred.

Example 2: Calculating Service Availability Metrics

The SLA is calculated every minute for the current minute, and is stored as a time-driven derived metric. When the Foglight® Management Server accepts forwarded data, the SLA is not re-calculated using the forwarded data.

For example, the SLA calculations take into consideration the expected average response time. The Foglight Management Server monitors a host that becomes disconnected for three hours due to a network outage. The host uses the Foglight Agent Manager to communicate with the server. At the time the connection is lost, the Foglight Agent Manager continues to collect the data from the host, storing the collected data on disk, including the collected data, which now contains the average response metrics.

The Foglight Management Server is configured to accept data that is older than one hour, with the store-and-forward time widow set to eight hours. The server’s configuration file, <foglight_home>/config/server.config, was edited prior to the server startup, and the option foglight.data_service.max_past_timestamp_delta is set to 28,800,000 milliseconds (which equals to eight hours) as follows:

The Agent Manager is aware of this configuration and continues to collect and store the data past the one-hour default, without deleting any new messages that come in.

During the connection disruption, the host’s average response metrics change over time. After three hours, the connection between the host and the server is restored, and the Foglight Agent Manager forwards the stored metrics to the server, including the average response metrics. The Foglight Management Server persists the forwarded data into the repository. The Foglight Management Server does not make use of the forwarded data to re-calculate the SLA.