How do I modify a working HBR configuration to use the more scalable HBREx session agent?

Answer

Contents

Problem

How do I modify a working HBR configuration to use the more scalable HBREx session agent?

Solution

The HBR session agent has some scalability issues that are removed by using HBREx in its place.? When using the HBR session agent all of the child pipelines share the same 32-bit memory space (limited to 2GB) as the main pipeline.? That is the total amount of memory used by all child pipelines and the main pipeline can be no more than 2 GB.? This places an effective limit on the number of child pipelines that can run on a single HBR regardless of how much memory or how many processors that machine may have.? The limit is based on what session agents are included in the main and child pipelines but with a typical pipeline configuration the maximum number of child
pipelines is under 20.

HBREx removes this limitation by allowing each child pipeline to run as a separate 32-bit process.? This allows the child pipelines to run on their own processors and each to use up to 2 GB of memory.? This allows the HBR hardware to be the limiting factor rather than architectural constraints

The process of moving an installation from using the HBR session agent to the HBREx session agent requires several changes both to the main pipeline configuration (TealeafCaptureSocket.cfg) and to the child pipeline configuration files.

Be sure to back up the working TealeafCaptureSocket.cfg and child pipeline configuration files before starting on this conversion.?

We recommend that the [HBR] section of TealeafCaptureSocket.cfg be copied and that the copy's section name be changed to "[HBREx]" so that common configuration
elements do not have to be recreated.? After all needed changes have been made to the main and child pipeline configurations the DownStreamConfigSection item that pointed to the HBR session agent section should be changed to point to the new HBREx section.

The following configuration items are significant in implementing the HBREx session agent:

• TypeName -- Set to "HBREx"
• DLL -- Set to "SessionAgentPipelineSplitterEx.dll"
• StartPortRangeChildPipelines -- Normally set to 2000.? This determines the starting port number that the HBREx session agent will use to send traffic to the child pipelines on the same machine.? The actual port used for the first child pipeline will be one greater than this setting i.e. 2001.? Thus an HBR with 16 child pipelines whose StartPortRangeChildPipelines? is set to 2000 will send traffic to ports 2001 through 2016.
• CanisterAppNameN -- Each CanisterAppName must be "TealeafCSS_" followed by that child pipeline's port number.? In the example above CanisterAppName1 would be "TealeafCSS_2001" and CanisterAppName16 would be "TealeafCSS_2016".
• AppName in [StatusMaster] section of the child pipeline configuration files -- This must be the same as that child pipeline's CanisterAppName as set in the [HBREx] section of the TealeafCaptureSocket.cfg file.
• LogDir in the [DecoupleEx] section of the child pipeline configuration files -- It is recommended that the DecoupleEx logs for each child pipeline be directed to a directory created just for that child pipeline.? This recommendation has resolved HBREx issues in the field but it is unclear why this may be needed as the files are named so that do not conflict with each other.

The message "Instance not yet determined" has been observed in the HBR_Poller log and Windows Application Event Log when the HBREx session agent is mis-configured or otherwise not working correctly.? The message appears to indicate that the CanisterAppNameN in the main pipeline config is not the same as the AppName in the parallel child pipeline config.

Applies to version(s): V 9.0.2