Release LoadGen Insight Agent: 1.0.0.12092

New Features

• Logon Duration Monitoring - The Insight Agent can now measure how long each user logon takes and break it down into its individual phases, so you can see where logon time is actually being spent. Enable it per machine group and set how long after logon the measurement is collected. Results are published as new userlogonmetric, userlogonphasemetric and userlogonprocessmetric measurements, giving you a per-logon summary, a per-phase timeline and an optional per-process breakdown for root-cause analysis. Phases are measured for standard Windows logons and, where present, are detected automatically for FSLogix, Citrix profile management and session startup, RDS and AVD, and Omnissa Horizon including DEM and App Volumes. Because phases are measured independently and can overlap, read them as a timeline rather than adding them up to the total logon time.
• Application Launch and Foreground Window Monitoring - The agent can now measure how long applications take to become usable after they are started, published as the new applicationlaunchmetric measurement, and track which application window a user is working in, published as the new foregroundwindowmetric measurement. Optional launch probes let you define what "ready" means for a given application, for example waiting until a specific control appears in the application window instead of the first window that opens. Requires process information monitoring to be enabled.

Enhancements

• Logon Monitoring Prerequisite Reporting - Every logon record reports whether the Windows event channels and audit settings the measurement depends on were actually available, so incomplete results are visible rather than silently wrong. An optional setting lets the agent configure those prerequisites for you.
• Expanded Linux Monitoring Coverage - Linux agents now also report service inventory as servicestatemetric, host, CPU and disk inventory as systemspecs, directory sizes as directorysizemetric, and closed network connections as processperformancemetricend_connection. Network throughput now uses real per-connection TCP byte and segment counters, matching what Windows agents report. The service and host inventory measurements are specific to Linux agents, so dashboards should query them by name.
• Monitoring Settings Now Applied on Linux - System, network, process and process information monitoring can each be enabled or disabled individually on Linux agents, and include and exclude filters are applied using the same matching rules as Windows agents. Previously Linux agents collected everything regardless of these settings.
• Consistent Agent Versioning on Linux - Linux agent builds and packages now use the same version scheme as Windows agents, so the Platform reports and compares agent versions correctly when deciding whether an update is needed.
• Hardened internal credential handling so that monitoring components receive only the settings they need

Bug Fixes

• Fixed Runaway Cache Growth on Windows Agents - The local cache database on Windows agents could grow to several gigabytes and never shrink, even while measurements were uploading successfully, because the routine that drains cached data was spending its time reclaiming disk space instead of sending. Cached data now drains reliably, and disk space is reclaimed only when there is a meaningful amount to release. The first service start after upgrading may take longer than usual while an existing large cache is optimized; later restarts are unaffected.
• Fixed Agent Updates on Packaged Linux Agents - Platform-initiated agent updates failed on Linux agents installed from a package, so those agents could not be updated from the Platform. Updates now apply correctly, and if any step fails the agent restores the previous version and restarts itself.
• Fixed Measurement Loss While InfluxDB Is Unavailable on Linux - Measurements collected while InfluxDB was unreachable were discarded. Linux agents now spool them to disk and re-send them, oldest first, once the destination recovers, following the existing cache database settings.
• Fixed Machine Group Settings Being Ignored on Linux - Tenant and machine group tags were not applied, and the check-in, upload and sampling intervals fell back to built-in defaults. These settings are now read correctly, and a configuration change pushed from the Platform takes effect without restarting the agent.
• Fixed Enrolled Linux Agents Not Starting Automatically - After installing or upgrading a package, an already-enrolled agent did not start on its own and had to be started manually.
• Fixed the Insight service losing buffered measurements when the service was stopped on Linux agents
• Fixed agent key updates so that a new key issued by the Platform is saved before it is used, preventing a restart from falling back to the previous key
• Improved the reliability of internal communication between the Control and Insight services on Linux agents

Looking ahead: Support for macOS agents is planned for Q4 2026.